2 * Copyright (C) 2005 - 2015 Emulex
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License version 2
7 * as published by the Free Software Foundation. The full GNU General
8 * Public License is included in this distribution in the file called COPYING.
10 * Contact Information:
11 * linux-drivers@emulex.com
15 * Costa Mesa, CA 92626
18 #include <linux/prefetch.h>
19 #include <linux/module.h>
22 #include <asm/div64.h>
23 #include <linux/aer.h>
24 #include <linux/if_bridge.h>
25 #include <net/busy_poll.h>
26 #include <net/vxlan.h>
28 MODULE_VERSION(DRV_VER
);
29 MODULE_DESCRIPTION(DRV_DESC
" " DRV_VER
);
30 MODULE_AUTHOR("Emulex Corporation");
31 MODULE_LICENSE("GPL");
33 /* num_vfs module param is obsolete.
34 * Use sysfs method to enable/disable VFs.
36 static unsigned int num_vfs
;
37 module_param(num_vfs
, uint
, S_IRUGO
);
38 MODULE_PARM_DESC(num_vfs
, "Number of PCI VFs to initialize");
40 static ushort rx_frag_size
= 2048;
41 module_param(rx_frag_size
, ushort
, S_IRUGO
);
42 MODULE_PARM_DESC(rx_frag_size
, "Size of a fragment that holds rcvd data.");
44 static const struct pci_device_id be_dev_ids
[] = {
45 { PCI_DEVICE(BE_VENDOR_ID
, BE_DEVICE_ID1
) },
46 { PCI_DEVICE(BE_VENDOR_ID
, BE_DEVICE_ID2
) },
47 { PCI_DEVICE(BE_VENDOR_ID
, OC_DEVICE_ID1
) },
48 { PCI_DEVICE(BE_VENDOR_ID
, OC_DEVICE_ID2
) },
49 { PCI_DEVICE(EMULEX_VENDOR_ID
, OC_DEVICE_ID3
)},
50 { PCI_DEVICE(EMULEX_VENDOR_ID
, OC_DEVICE_ID4
)},
51 { PCI_DEVICE(EMULEX_VENDOR_ID
, OC_DEVICE_ID5
)},
52 { PCI_DEVICE(EMULEX_VENDOR_ID
, OC_DEVICE_ID6
)},
55 MODULE_DEVICE_TABLE(pci
, be_dev_ids
);
56 /* UE Status Low CSR */
57 static const char * const ue_status_low_desc
[] = {
92 /* UE Status High CSR */
93 static const char * const ue_status_hi_desc
[] = {
128 #define BE_VF_IF_EN_FLAGS (BE_IF_FLAGS_UNTAGGED | \
129 BE_IF_FLAGS_BROADCAST | \
130 BE_IF_FLAGS_MULTICAST | \
131 BE_IF_FLAGS_PASS_L3L4_ERRORS)
133 static void be_queue_free(struct be_adapter
*adapter
, struct be_queue_info
*q
)
135 struct be_dma_mem
*mem
= &q
->dma_mem
;
138 dma_free_coherent(&adapter
->pdev
->dev
, mem
->size
, mem
->va
,
144 static int be_queue_alloc(struct be_adapter
*adapter
, struct be_queue_info
*q
,
145 u16 len
, u16 entry_size
)
147 struct be_dma_mem
*mem
= &q
->dma_mem
;
149 memset(q
, 0, sizeof(*q
));
151 q
->entry_size
= entry_size
;
152 mem
->size
= len
* entry_size
;
153 mem
->va
= dma_zalloc_coherent(&adapter
->pdev
->dev
, mem
->size
, &mem
->dma
,
160 static void be_reg_intr_set(struct be_adapter
*adapter
, bool enable
)
164 pci_read_config_dword(adapter
->pdev
, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET
,
166 enabled
= reg
& MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK
;
168 if (!enabled
&& enable
)
169 reg
|= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK
;
170 else if (enabled
&& !enable
)
171 reg
&= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK
;
175 pci_write_config_dword(adapter
->pdev
,
176 PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET
, reg
);
179 static void be_intr_set(struct be_adapter
*adapter
, bool enable
)
183 /* On lancer interrupts can't be controlled via this register */
184 if (lancer_chip(adapter
))
187 if (be_check_error(adapter
, BE_ERROR_EEH
))
190 status
= be_cmd_intr_set(adapter
, enable
);
192 be_reg_intr_set(adapter
, enable
);
195 static void be_rxq_notify(struct be_adapter
*adapter
, u16 qid
, u16 posted
)
199 if (be_check_error(adapter
, BE_ERROR_HW
))
202 val
|= qid
& DB_RQ_RING_ID_MASK
;
203 val
|= posted
<< DB_RQ_NUM_POSTED_SHIFT
;
206 iowrite32(val
, adapter
->db
+ DB_RQ_OFFSET
);
209 static void be_txq_notify(struct be_adapter
*adapter
, struct be_tx_obj
*txo
,
214 if (be_check_error(adapter
, BE_ERROR_HW
))
217 val
|= txo
->q
.id
& DB_TXULP_RING_ID_MASK
;
218 val
|= (posted
& DB_TXULP_NUM_POSTED_MASK
) << DB_TXULP_NUM_POSTED_SHIFT
;
221 iowrite32(val
, adapter
->db
+ txo
->db_offset
);
224 static void be_eq_notify(struct be_adapter
*adapter
, u16 qid
,
225 bool arm
, bool clear_int
, u16 num_popped
,
226 u32 eq_delay_mult_enc
)
230 val
|= qid
& DB_EQ_RING_ID_MASK
;
231 val
|= ((qid
& DB_EQ_RING_ID_EXT_MASK
) << DB_EQ_RING_ID_EXT_MASK_SHIFT
);
233 if (be_check_error(adapter
, BE_ERROR_HW
))
237 val
|= 1 << DB_EQ_REARM_SHIFT
;
239 val
|= 1 << DB_EQ_CLR_SHIFT
;
240 val
|= 1 << DB_EQ_EVNT_SHIFT
;
241 val
|= num_popped
<< DB_EQ_NUM_POPPED_SHIFT
;
242 val
|= eq_delay_mult_enc
<< DB_EQ_R2I_DLY_SHIFT
;
243 iowrite32(val
, adapter
->db
+ DB_EQ_OFFSET
);
246 void be_cq_notify(struct be_adapter
*adapter
, u16 qid
, bool arm
, u16 num_popped
)
250 val
|= qid
& DB_CQ_RING_ID_MASK
;
251 val
|= ((qid
& DB_CQ_RING_ID_EXT_MASK
) <<
252 DB_CQ_RING_ID_EXT_MASK_SHIFT
);
254 if (be_check_error(adapter
, BE_ERROR_HW
))
258 val
|= 1 << DB_CQ_REARM_SHIFT
;
259 val
|= num_popped
<< DB_CQ_NUM_POPPED_SHIFT
;
260 iowrite32(val
, adapter
->db
+ DB_CQ_OFFSET
);
263 static int be_mac_addr_set(struct net_device
*netdev
, void *p
)
265 struct be_adapter
*adapter
= netdev_priv(netdev
);
266 struct device
*dev
= &adapter
->pdev
->dev
;
267 struct sockaddr
*addr
= p
;
270 u32 old_pmac_id
= adapter
->pmac_id
[0], curr_pmac_id
= 0;
272 if (!is_valid_ether_addr(addr
->sa_data
))
273 return -EADDRNOTAVAIL
;
275 /* Proceed further only if, User provided MAC is different
278 if (ether_addr_equal(addr
->sa_data
, netdev
->dev_addr
))
281 /* if device is not running, copy MAC to netdev->dev_addr */
282 if (!netif_running(netdev
))
285 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
286 * privilege or if PF did not provision the new MAC address.
287 * On BE3, this cmd will always fail if the VF doesn't have the
288 * FILTMGMT privilege. This failure is OK, only if the PF programmed
289 * the MAC for the VF.
291 status
= be_cmd_pmac_add(adapter
, (u8
*)addr
->sa_data
,
292 adapter
->if_handle
, &adapter
->pmac_id
[0], 0);
294 curr_pmac_id
= adapter
->pmac_id
[0];
296 /* Delete the old programmed MAC. This call may fail if the
297 * old MAC was already deleted by the PF driver.
299 if (adapter
->pmac_id
[0] != old_pmac_id
)
300 be_cmd_pmac_del(adapter
, adapter
->if_handle
,
304 /* Decide if the new MAC is successfully activated only after
307 status
= be_cmd_get_active_mac(adapter
, curr_pmac_id
, mac
,
308 adapter
->if_handle
, true, 0);
312 /* The MAC change did not happen, either due to lack of privilege
313 * or PF didn't pre-provision.
315 if (!ether_addr_equal(addr
->sa_data
, mac
)) {
320 ether_addr_copy(netdev
->dev_addr
, addr
->sa_data
);
321 dev_info(dev
, "MAC address changed to %pM\n", addr
->sa_data
);
324 dev_warn(dev
, "MAC address change to %pM failed\n", addr
->sa_data
);
328 /* BE2 supports only v0 cmd */
329 static void *hw_stats_from_cmd(struct be_adapter
*adapter
)
331 if (BE2_chip(adapter
)) {
332 struct be_cmd_resp_get_stats_v0
*cmd
= adapter
->stats_cmd
.va
;
334 return &cmd
->hw_stats
;
335 } else if (BE3_chip(adapter
)) {
336 struct be_cmd_resp_get_stats_v1
*cmd
= adapter
->stats_cmd
.va
;
338 return &cmd
->hw_stats
;
340 struct be_cmd_resp_get_stats_v2
*cmd
= adapter
->stats_cmd
.va
;
342 return &cmd
->hw_stats
;
346 /* BE2 supports only v0 cmd */
347 static void *be_erx_stats_from_cmd(struct be_adapter
*adapter
)
349 if (BE2_chip(adapter
)) {
350 struct be_hw_stats_v0
*hw_stats
= hw_stats_from_cmd(adapter
);
352 return &hw_stats
->erx
;
353 } else if (BE3_chip(adapter
)) {
354 struct be_hw_stats_v1
*hw_stats
= hw_stats_from_cmd(adapter
);
356 return &hw_stats
->erx
;
358 struct be_hw_stats_v2
*hw_stats
= hw_stats_from_cmd(adapter
);
360 return &hw_stats
->erx
;
364 static void populate_be_v0_stats(struct be_adapter
*adapter
)
366 struct be_hw_stats_v0
*hw_stats
= hw_stats_from_cmd(adapter
);
367 struct be_pmem_stats
*pmem_sts
= &hw_stats
->pmem
;
368 struct be_rxf_stats_v0
*rxf_stats
= &hw_stats
->rxf
;
369 struct be_port_rxf_stats_v0
*port_stats
=
370 &rxf_stats
->port
[adapter
->port_num
];
371 struct be_drv_stats
*drvs
= &adapter
->drv_stats
;
373 be_dws_le_to_cpu(hw_stats
, sizeof(*hw_stats
));
374 drvs
->rx_pause_frames
= port_stats
->rx_pause_frames
;
375 drvs
->rx_crc_errors
= port_stats
->rx_crc_errors
;
376 drvs
->rx_control_frames
= port_stats
->rx_control_frames
;
377 drvs
->rx_in_range_errors
= port_stats
->rx_in_range_errors
;
378 drvs
->rx_frame_too_long
= port_stats
->rx_frame_too_long
;
379 drvs
->rx_dropped_runt
= port_stats
->rx_dropped_runt
;
380 drvs
->rx_ip_checksum_errs
= port_stats
->rx_ip_checksum_errs
;
381 drvs
->rx_tcp_checksum_errs
= port_stats
->rx_tcp_checksum_errs
;
382 drvs
->rx_udp_checksum_errs
= port_stats
->rx_udp_checksum_errs
;
383 drvs
->rxpp_fifo_overflow_drop
= port_stats
->rx_fifo_overflow
;
384 drvs
->rx_dropped_tcp_length
= port_stats
->rx_dropped_tcp_length
;
385 drvs
->rx_dropped_too_small
= port_stats
->rx_dropped_too_small
;
386 drvs
->rx_dropped_too_short
= port_stats
->rx_dropped_too_short
;
387 drvs
->rx_out_range_errors
= port_stats
->rx_out_range_errors
;
388 drvs
->rx_input_fifo_overflow_drop
= port_stats
->rx_input_fifo_overflow
;
389 drvs
->rx_dropped_header_too_small
=
390 port_stats
->rx_dropped_header_too_small
;
391 drvs
->rx_address_filtered
=
392 port_stats
->rx_address_filtered
+
393 port_stats
->rx_vlan_filtered
;
394 drvs
->rx_alignment_symbol_errors
=
395 port_stats
->rx_alignment_symbol_errors
;
397 drvs
->tx_pauseframes
= port_stats
->tx_pauseframes
;
398 drvs
->tx_controlframes
= port_stats
->tx_controlframes
;
400 if (adapter
->port_num
)
401 drvs
->jabber_events
= rxf_stats
->port1_jabber_events
;
403 drvs
->jabber_events
= rxf_stats
->port0_jabber_events
;
404 drvs
->rx_drops_no_pbuf
= rxf_stats
->rx_drops_no_pbuf
;
405 drvs
->rx_drops_no_erx_descr
= rxf_stats
->rx_drops_no_erx_descr
;
406 drvs
->forwarded_packets
= rxf_stats
->forwarded_packets
;
407 drvs
->rx_drops_mtu
= rxf_stats
->rx_drops_mtu
;
408 drvs
->rx_drops_no_tpre_descr
= rxf_stats
->rx_drops_no_tpre_descr
;
409 drvs
->rx_drops_too_many_frags
= rxf_stats
->rx_drops_too_many_frags
;
410 adapter
->drv_stats
.eth_red_drops
= pmem_sts
->eth_red_drops
;
413 static void populate_be_v1_stats(struct be_adapter
*adapter
)
415 struct be_hw_stats_v1
*hw_stats
= hw_stats_from_cmd(adapter
);
416 struct be_pmem_stats
*pmem_sts
= &hw_stats
->pmem
;
417 struct be_rxf_stats_v1
*rxf_stats
= &hw_stats
->rxf
;
418 struct be_port_rxf_stats_v1
*port_stats
=
419 &rxf_stats
->port
[adapter
->port_num
];
420 struct be_drv_stats
*drvs
= &adapter
->drv_stats
;
422 be_dws_le_to_cpu(hw_stats
, sizeof(*hw_stats
));
423 drvs
->pmem_fifo_overflow_drop
= port_stats
->pmem_fifo_overflow_drop
;
424 drvs
->rx_priority_pause_frames
= port_stats
->rx_priority_pause_frames
;
425 drvs
->rx_pause_frames
= port_stats
->rx_pause_frames
;
426 drvs
->rx_crc_errors
= port_stats
->rx_crc_errors
;
427 drvs
->rx_control_frames
= port_stats
->rx_control_frames
;
428 drvs
->rx_in_range_errors
= port_stats
->rx_in_range_errors
;
429 drvs
->rx_frame_too_long
= port_stats
->rx_frame_too_long
;
430 drvs
->rx_dropped_runt
= port_stats
->rx_dropped_runt
;
431 drvs
->rx_ip_checksum_errs
= port_stats
->rx_ip_checksum_errs
;
432 drvs
->rx_tcp_checksum_errs
= port_stats
->rx_tcp_checksum_errs
;
433 drvs
->rx_udp_checksum_errs
= port_stats
->rx_udp_checksum_errs
;
434 drvs
->rx_dropped_tcp_length
= port_stats
->rx_dropped_tcp_length
;
435 drvs
->rx_dropped_too_small
= port_stats
->rx_dropped_too_small
;
436 drvs
->rx_dropped_too_short
= port_stats
->rx_dropped_too_short
;
437 drvs
->rx_out_range_errors
= port_stats
->rx_out_range_errors
;
438 drvs
->rx_dropped_header_too_small
=
439 port_stats
->rx_dropped_header_too_small
;
440 drvs
->rx_input_fifo_overflow_drop
=
441 port_stats
->rx_input_fifo_overflow_drop
;
442 drvs
->rx_address_filtered
= port_stats
->rx_address_filtered
;
443 drvs
->rx_alignment_symbol_errors
=
444 port_stats
->rx_alignment_symbol_errors
;
445 drvs
->rxpp_fifo_overflow_drop
= port_stats
->rxpp_fifo_overflow_drop
;
446 drvs
->tx_pauseframes
= port_stats
->tx_pauseframes
;
447 drvs
->tx_controlframes
= port_stats
->tx_controlframes
;
448 drvs
->tx_priority_pauseframes
= port_stats
->tx_priority_pauseframes
;
449 drvs
->jabber_events
= port_stats
->jabber_events
;
450 drvs
->rx_drops_no_pbuf
= rxf_stats
->rx_drops_no_pbuf
;
451 drvs
->rx_drops_no_erx_descr
= rxf_stats
->rx_drops_no_erx_descr
;
452 drvs
->forwarded_packets
= rxf_stats
->forwarded_packets
;
453 drvs
->rx_drops_mtu
= rxf_stats
->rx_drops_mtu
;
454 drvs
->rx_drops_no_tpre_descr
= rxf_stats
->rx_drops_no_tpre_descr
;
455 drvs
->rx_drops_too_many_frags
= rxf_stats
->rx_drops_too_many_frags
;
456 adapter
->drv_stats
.eth_red_drops
= pmem_sts
->eth_red_drops
;
459 static void populate_be_v2_stats(struct be_adapter
*adapter
)
461 struct be_hw_stats_v2
*hw_stats
= hw_stats_from_cmd(adapter
);
462 struct be_pmem_stats
*pmem_sts
= &hw_stats
->pmem
;
463 struct be_rxf_stats_v2
*rxf_stats
= &hw_stats
->rxf
;
464 struct be_port_rxf_stats_v2
*port_stats
=
465 &rxf_stats
->port
[adapter
->port_num
];
466 struct be_drv_stats
*drvs
= &adapter
->drv_stats
;
468 be_dws_le_to_cpu(hw_stats
, sizeof(*hw_stats
));
469 drvs
->pmem_fifo_overflow_drop
= port_stats
->pmem_fifo_overflow_drop
;
470 drvs
->rx_priority_pause_frames
= port_stats
->rx_priority_pause_frames
;
471 drvs
->rx_pause_frames
= port_stats
->rx_pause_frames
;
472 drvs
->rx_crc_errors
= port_stats
->rx_crc_errors
;
473 drvs
->rx_control_frames
= port_stats
->rx_control_frames
;
474 drvs
->rx_in_range_errors
= port_stats
->rx_in_range_errors
;
475 drvs
->rx_frame_too_long
= port_stats
->rx_frame_too_long
;
476 drvs
->rx_dropped_runt
= port_stats
->rx_dropped_runt
;
477 drvs
->rx_ip_checksum_errs
= port_stats
->rx_ip_checksum_errs
;
478 drvs
->rx_tcp_checksum_errs
= port_stats
->rx_tcp_checksum_errs
;
479 drvs
->rx_udp_checksum_errs
= port_stats
->rx_udp_checksum_errs
;
480 drvs
->rx_dropped_tcp_length
= port_stats
->rx_dropped_tcp_length
;
481 drvs
->rx_dropped_too_small
= port_stats
->rx_dropped_too_small
;
482 drvs
->rx_dropped_too_short
= port_stats
->rx_dropped_too_short
;
483 drvs
->rx_out_range_errors
= port_stats
->rx_out_range_errors
;
484 drvs
->rx_dropped_header_too_small
=
485 port_stats
->rx_dropped_header_too_small
;
486 drvs
->rx_input_fifo_overflow_drop
=
487 port_stats
->rx_input_fifo_overflow_drop
;
488 drvs
->rx_address_filtered
= port_stats
->rx_address_filtered
;
489 drvs
->rx_alignment_symbol_errors
=
490 port_stats
->rx_alignment_symbol_errors
;
491 drvs
->rxpp_fifo_overflow_drop
= port_stats
->rxpp_fifo_overflow_drop
;
492 drvs
->tx_pauseframes
= port_stats
->tx_pauseframes
;
493 drvs
->tx_controlframes
= port_stats
->tx_controlframes
;
494 drvs
->tx_priority_pauseframes
= port_stats
->tx_priority_pauseframes
;
495 drvs
->jabber_events
= port_stats
->jabber_events
;
496 drvs
->rx_drops_no_pbuf
= rxf_stats
->rx_drops_no_pbuf
;
497 drvs
->rx_drops_no_erx_descr
= rxf_stats
->rx_drops_no_erx_descr
;
498 drvs
->forwarded_packets
= rxf_stats
->forwarded_packets
;
499 drvs
->rx_drops_mtu
= rxf_stats
->rx_drops_mtu
;
500 drvs
->rx_drops_no_tpre_descr
= rxf_stats
->rx_drops_no_tpre_descr
;
501 drvs
->rx_drops_too_many_frags
= rxf_stats
->rx_drops_too_many_frags
;
502 adapter
->drv_stats
.eth_red_drops
= pmem_sts
->eth_red_drops
;
503 if (be_roce_supported(adapter
)) {
504 drvs
->rx_roce_bytes_lsd
= port_stats
->roce_bytes_received_lsd
;
505 drvs
->rx_roce_bytes_msd
= port_stats
->roce_bytes_received_msd
;
506 drvs
->rx_roce_frames
= port_stats
->roce_frames_received
;
507 drvs
->roce_drops_crc
= port_stats
->roce_drops_crc
;
508 drvs
->roce_drops_payload_len
=
509 port_stats
->roce_drops_payload_len
;
513 static void populate_lancer_stats(struct be_adapter
*adapter
)
515 struct be_drv_stats
*drvs
= &adapter
->drv_stats
;
516 struct lancer_pport_stats
*pport_stats
= pport_stats_from_cmd(adapter
);
518 be_dws_le_to_cpu(pport_stats
, sizeof(*pport_stats
));
519 drvs
->rx_pause_frames
= pport_stats
->rx_pause_frames_lo
;
520 drvs
->rx_crc_errors
= pport_stats
->rx_crc_errors_lo
;
521 drvs
->rx_control_frames
= pport_stats
->rx_control_frames_lo
;
522 drvs
->rx_in_range_errors
= pport_stats
->rx_in_range_errors
;
523 drvs
->rx_frame_too_long
= pport_stats
->rx_frames_too_long_lo
;
524 drvs
->rx_dropped_runt
= pport_stats
->rx_dropped_runt
;
525 drvs
->rx_ip_checksum_errs
= pport_stats
->rx_ip_checksum_errors
;
526 drvs
->rx_tcp_checksum_errs
= pport_stats
->rx_tcp_checksum_errors
;
527 drvs
->rx_udp_checksum_errs
= pport_stats
->rx_udp_checksum_errors
;
528 drvs
->rx_dropped_tcp_length
=
529 pport_stats
->rx_dropped_invalid_tcp_length
;
530 drvs
->rx_dropped_too_small
= pport_stats
->rx_dropped_too_small
;
531 drvs
->rx_dropped_too_short
= pport_stats
->rx_dropped_too_short
;
532 drvs
->rx_out_range_errors
= pport_stats
->rx_out_of_range_errors
;
533 drvs
->rx_dropped_header_too_small
=
534 pport_stats
->rx_dropped_header_too_small
;
535 drvs
->rx_input_fifo_overflow_drop
= pport_stats
->rx_fifo_overflow
;
536 drvs
->rx_address_filtered
=
537 pport_stats
->rx_address_filtered
+
538 pport_stats
->rx_vlan_filtered
;
539 drvs
->rx_alignment_symbol_errors
= pport_stats
->rx_symbol_errors_lo
;
540 drvs
->rxpp_fifo_overflow_drop
= pport_stats
->rx_fifo_overflow
;
541 drvs
->tx_pauseframes
= pport_stats
->tx_pause_frames_lo
;
542 drvs
->tx_controlframes
= pport_stats
->tx_control_frames_lo
;
543 drvs
->jabber_events
= pport_stats
->rx_jabbers
;
544 drvs
->forwarded_packets
= pport_stats
->num_forwards_lo
;
545 drvs
->rx_drops_mtu
= pport_stats
->rx_drops_mtu_lo
;
546 drvs
->rx_drops_too_many_frags
=
547 pport_stats
->rx_drops_too_many_frags_lo
;
550 static void accumulate_16bit_val(u32
*acc
, u16 val
)
552 #define lo(x) (x & 0xFFFF)
553 #define hi(x) (x & 0xFFFF0000)
554 bool wrapped
= val
< lo(*acc
);
555 u32 newacc
= hi(*acc
) + val
;
559 ACCESS_ONCE(*acc
) = newacc
;
562 static void populate_erx_stats(struct be_adapter
*adapter
,
563 struct be_rx_obj
*rxo
, u32 erx_stat
)
565 if (!BEx_chip(adapter
))
566 rx_stats(rxo
)->rx_drops_no_frags
= erx_stat
;
568 /* below erx HW counter can actually wrap around after
569 * 65535. Driver accumulates a 32-bit value
571 accumulate_16bit_val(&rx_stats(rxo
)->rx_drops_no_frags
,
575 void be_parse_stats(struct be_adapter
*adapter
)
577 struct be_erx_stats_v2
*erx
= be_erx_stats_from_cmd(adapter
);
578 struct be_rx_obj
*rxo
;
582 if (lancer_chip(adapter
)) {
583 populate_lancer_stats(adapter
);
585 if (BE2_chip(adapter
))
586 populate_be_v0_stats(adapter
);
587 else if (BE3_chip(adapter
))
589 populate_be_v1_stats(adapter
);
591 populate_be_v2_stats(adapter
);
593 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
594 for_all_rx_queues(adapter
, rxo
, i
) {
595 erx_stat
= erx
->rx_drops_no_fragments
[rxo
->q
.id
];
596 populate_erx_stats(adapter
, rxo
, erx_stat
);
601 static struct rtnl_link_stats64
*be_get_stats64(struct net_device
*netdev
,
602 struct rtnl_link_stats64
*stats
)
604 struct be_adapter
*adapter
= netdev_priv(netdev
);
605 struct be_drv_stats
*drvs
= &adapter
->drv_stats
;
606 struct be_rx_obj
*rxo
;
607 struct be_tx_obj
*txo
;
612 for_all_rx_queues(adapter
, rxo
, i
) {
613 const struct be_rx_stats
*rx_stats
= rx_stats(rxo
);
616 start
= u64_stats_fetch_begin_irq(&rx_stats
->sync
);
617 pkts
= rx_stats(rxo
)->rx_pkts
;
618 bytes
= rx_stats(rxo
)->rx_bytes
;
619 } while (u64_stats_fetch_retry_irq(&rx_stats
->sync
, start
));
620 stats
->rx_packets
+= pkts
;
621 stats
->rx_bytes
+= bytes
;
622 stats
->multicast
+= rx_stats(rxo
)->rx_mcast_pkts
;
623 stats
->rx_dropped
+= rx_stats(rxo
)->rx_drops_no_skbs
+
624 rx_stats(rxo
)->rx_drops_no_frags
;
627 for_all_tx_queues(adapter
, txo
, i
) {
628 const struct be_tx_stats
*tx_stats
= tx_stats(txo
);
631 start
= u64_stats_fetch_begin_irq(&tx_stats
->sync
);
632 pkts
= tx_stats(txo
)->tx_pkts
;
633 bytes
= tx_stats(txo
)->tx_bytes
;
634 } while (u64_stats_fetch_retry_irq(&tx_stats
->sync
, start
));
635 stats
->tx_packets
+= pkts
;
636 stats
->tx_bytes
+= bytes
;
639 /* bad pkts received */
640 stats
->rx_errors
= drvs
->rx_crc_errors
+
641 drvs
->rx_alignment_symbol_errors
+
642 drvs
->rx_in_range_errors
+
643 drvs
->rx_out_range_errors
+
644 drvs
->rx_frame_too_long
+
645 drvs
->rx_dropped_too_small
+
646 drvs
->rx_dropped_too_short
+
647 drvs
->rx_dropped_header_too_small
+
648 drvs
->rx_dropped_tcp_length
+
649 drvs
->rx_dropped_runt
;
651 /* detailed rx errors */
652 stats
->rx_length_errors
= drvs
->rx_in_range_errors
+
653 drvs
->rx_out_range_errors
+
654 drvs
->rx_frame_too_long
;
656 stats
->rx_crc_errors
= drvs
->rx_crc_errors
;
658 /* frame alignment errors */
659 stats
->rx_frame_errors
= drvs
->rx_alignment_symbol_errors
;
661 /* receiver fifo overrun */
662 /* drops_no_pbuf is no per i/f, it's per BE card */
663 stats
->rx_fifo_errors
= drvs
->rxpp_fifo_overflow_drop
+
664 drvs
->rx_input_fifo_overflow_drop
+
665 drvs
->rx_drops_no_pbuf
;
669 void be_link_status_update(struct be_adapter
*adapter
, u8 link_status
)
671 struct net_device
*netdev
= adapter
->netdev
;
673 if (!(adapter
->flags
& BE_FLAGS_LINK_STATUS_INIT
)) {
674 netif_carrier_off(netdev
);
675 adapter
->flags
|= BE_FLAGS_LINK_STATUS_INIT
;
679 netif_carrier_on(netdev
);
681 netif_carrier_off(netdev
);
683 netdev_info(netdev
, "Link is %s\n", link_status
? "Up" : "Down");
686 static void be_tx_stats_update(struct be_tx_obj
*txo
, struct sk_buff
*skb
)
688 struct be_tx_stats
*stats
= tx_stats(txo
);
689 u64 tx_pkts
= skb_shinfo(skb
)->gso_segs
? : 1;
691 u64_stats_update_begin(&stats
->sync
);
693 stats
->tx_bytes
+= skb
->len
;
694 stats
->tx_pkts
+= tx_pkts
;
695 if (skb
->encapsulation
&& skb
->ip_summed
== CHECKSUM_PARTIAL
)
696 stats
->tx_vxlan_offload_pkts
+= tx_pkts
;
697 u64_stats_update_end(&stats
->sync
);
700 /* Returns number of WRBs needed for the skb */
701 static u32
skb_wrb_cnt(struct sk_buff
*skb
)
703 /* +1 for the header wrb */
704 return 1 + (skb_headlen(skb
) ? 1 : 0) + skb_shinfo(skb
)->nr_frags
;
707 static inline void wrb_fill(struct be_eth_wrb
*wrb
, u64 addr
, int len
)
709 wrb
->frag_pa_hi
= cpu_to_le32(upper_32_bits(addr
));
710 wrb
->frag_pa_lo
= cpu_to_le32(lower_32_bits(addr
));
711 wrb
->frag_len
= cpu_to_le32(len
& ETH_WRB_FRAG_LEN_MASK
);
715 /* A dummy wrb is just all zeros. Using a separate routine for dummy-wrb
716 * to avoid the swap and shift/mask operations in wrb_fill().
718 static inline void wrb_fill_dummy(struct be_eth_wrb
*wrb
)
726 static inline u16
be_get_tx_vlan_tag(struct be_adapter
*adapter
,
732 vlan_tag
= skb_vlan_tag_get(skb
);
733 vlan_prio
= (vlan_tag
& VLAN_PRIO_MASK
) >> VLAN_PRIO_SHIFT
;
734 /* If vlan priority provided by OS is NOT in available bmap */
735 if (!(adapter
->vlan_prio_bmap
& (1 << vlan_prio
)))
736 vlan_tag
= (vlan_tag
& ~VLAN_PRIO_MASK
) |
737 adapter
->recommended_prio_bits
;
742 /* Used only for IP tunnel packets */
743 static u16
skb_inner_ip_proto(struct sk_buff
*skb
)
745 return (inner_ip_hdr(skb
)->version
== 4) ?
746 inner_ip_hdr(skb
)->protocol
: inner_ipv6_hdr(skb
)->nexthdr
;
749 static u16
skb_ip_proto(struct sk_buff
*skb
)
751 return (ip_hdr(skb
)->version
== 4) ?
752 ip_hdr(skb
)->protocol
: ipv6_hdr(skb
)->nexthdr
;
755 static inline bool be_is_txq_full(struct be_tx_obj
*txo
)
757 return atomic_read(&txo
->q
.used
) + BE_MAX_TX_FRAG_COUNT
>= txo
->q
.len
;
760 static inline bool be_can_txq_wake(struct be_tx_obj
*txo
)
762 return atomic_read(&txo
->q
.used
) < txo
->q
.len
/ 2;
765 static inline bool be_is_tx_compl_pending(struct be_tx_obj
*txo
)
767 return atomic_read(&txo
->q
.used
) > txo
->pend_wrb_cnt
;
770 static void be_get_wrb_params_from_skb(struct be_adapter
*adapter
,
772 struct be_wrb_params
*wrb_params
)
776 if (skb_is_gso(skb
)) {
777 BE_WRB_F_SET(wrb_params
->features
, LSO
, 1);
778 wrb_params
->lso_mss
= skb_shinfo(skb
)->gso_size
;
779 if (skb_is_gso_v6(skb
) && !lancer_chip(adapter
))
780 BE_WRB_F_SET(wrb_params
->features
, LSO6
, 1);
781 } else if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
782 if (skb
->encapsulation
) {
783 BE_WRB_F_SET(wrb_params
->features
, IPCS
, 1);
784 proto
= skb_inner_ip_proto(skb
);
786 proto
= skb_ip_proto(skb
);
788 if (proto
== IPPROTO_TCP
)
789 BE_WRB_F_SET(wrb_params
->features
, TCPCS
, 1);
790 else if (proto
== IPPROTO_UDP
)
791 BE_WRB_F_SET(wrb_params
->features
, UDPCS
, 1);
794 if (skb_vlan_tag_present(skb
)) {
795 BE_WRB_F_SET(wrb_params
->features
, VLAN
, 1);
796 wrb_params
->vlan_tag
= be_get_tx_vlan_tag(adapter
, skb
);
799 BE_WRB_F_SET(wrb_params
->features
, CRC
, 1);
802 static void wrb_fill_hdr(struct be_adapter
*adapter
,
803 struct be_eth_hdr_wrb
*hdr
,
804 struct be_wrb_params
*wrb_params
,
807 memset(hdr
, 0, sizeof(*hdr
));
809 SET_TX_WRB_HDR_BITS(crc
, hdr
,
810 BE_WRB_F_GET(wrb_params
->features
, CRC
));
811 SET_TX_WRB_HDR_BITS(ipcs
, hdr
,
812 BE_WRB_F_GET(wrb_params
->features
, IPCS
));
813 SET_TX_WRB_HDR_BITS(tcpcs
, hdr
,
814 BE_WRB_F_GET(wrb_params
->features
, TCPCS
));
815 SET_TX_WRB_HDR_BITS(udpcs
, hdr
,
816 BE_WRB_F_GET(wrb_params
->features
, UDPCS
));
818 SET_TX_WRB_HDR_BITS(lso
, hdr
,
819 BE_WRB_F_GET(wrb_params
->features
, LSO
));
820 SET_TX_WRB_HDR_BITS(lso6
, hdr
,
821 BE_WRB_F_GET(wrb_params
->features
, LSO6
));
822 SET_TX_WRB_HDR_BITS(lso_mss
, hdr
, wrb_params
->lso_mss
);
824 /* Hack to skip HW VLAN tagging needs evt = 1, compl = 0. When this
825 * hack is not needed, the evt bit is set while ringing DB.
827 SET_TX_WRB_HDR_BITS(event
, hdr
,
828 BE_WRB_F_GET(wrb_params
->features
, VLAN_SKIP_HW
));
829 SET_TX_WRB_HDR_BITS(vlan
, hdr
,
830 BE_WRB_F_GET(wrb_params
->features
, VLAN
));
831 SET_TX_WRB_HDR_BITS(vlan_tag
, hdr
, wrb_params
->vlan_tag
);
833 SET_TX_WRB_HDR_BITS(num_wrb
, hdr
, skb_wrb_cnt(skb
));
834 SET_TX_WRB_HDR_BITS(len
, hdr
, skb
->len
);
835 SET_TX_WRB_HDR_BITS(mgmt
, hdr
,
836 BE_WRB_F_GET(wrb_params
->features
, OS2BMC
));
839 static void unmap_tx_frag(struct device
*dev
, struct be_eth_wrb
*wrb
,
843 u32 frag_len
= le32_to_cpu(wrb
->frag_len
);
846 dma
= (u64
)le32_to_cpu(wrb
->frag_pa_hi
) << 32 |
847 (u64
)le32_to_cpu(wrb
->frag_pa_lo
);
850 dma_unmap_single(dev
, dma
, frag_len
, DMA_TO_DEVICE
);
852 dma_unmap_page(dev
, dma
, frag_len
, DMA_TO_DEVICE
);
856 /* Grab a WRB header for xmit */
857 static u32
be_tx_get_wrb_hdr(struct be_tx_obj
*txo
)
859 u32 head
= txo
->q
.head
;
861 queue_head_inc(&txo
->q
);
865 /* Set up the WRB header for xmit */
866 static void be_tx_setup_wrb_hdr(struct be_adapter
*adapter
,
867 struct be_tx_obj
*txo
,
868 struct be_wrb_params
*wrb_params
,
869 struct sk_buff
*skb
, u16 head
)
871 u32 num_frags
= skb_wrb_cnt(skb
);
872 struct be_queue_info
*txq
= &txo
->q
;
873 struct be_eth_hdr_wrb
*hdr
= queue_index_node(txq
, head
);
875 wrb_fill_hdr(adapter
, hdr
, wrb_params
, skb
);
876 be_dws_cpu_to_le(hdr
, sizeof(*hdr
));
878 BUG_ON(txo
->sent_skb_list
[head
]);
879 txo
->sent_skb_list
[head
] = skb
;
880 txo
->last_req_hdr
= head
;
881 atomic_add(num_frags
, &txq
->used
);
882 txo
->last_req_wrb_cnt
= num_frags
;
883 txo
->pend_wrb_cnt
+= num_frags
;
886 /* Setup a WRB fragment (buffer descriptor) for xmit */
887 static void be_tx_setup_wrb_frag(struct be_tx_obj
*txo
, dma_addr_t busaddr
,
890 struct be_eth_wrb
*wrb
;
891 struct be_queue_info
*txq
= &txo
->q
;
893 wrb
= queue_head_node(txq
);
894 wrb_fill(wrb
, busaddr
, len
);
898 /* Bring the queue back to the state it was in before be_xmit_enqueue() routine
899 * was invoked. The producer index is restored to the previous packet and the
900 * WRBs of the current packet are unmapped. Invoked to handle tx setup errors.
902 static void be_xmit_restore(struct be_adapter
*adapter
,
903 struct be_tx_obj
*txo
, u32 head
, bool map_single
,
907 struct be_eth_wrb
*wrb
;
908 struct be_queue_info
*txq
= &txo
->q
;
910 dev
= &adapter
->pdev
->dev
;
913 /* skip the first wrb (hdr); it's not mapped */
916 wrb
= queue_head_node(txq
);
917 unmap_tx_frag(dev
, wrb
, map_single
);
919 copied
-= le32_to_cpu(wrb
->frag_len
);
926 /* Enqueue the given packet for transmit. This routine allocates WRBs for the
927 * packet, dma maps the packet buffers and sets up the WRBs. Returns the number
928 * of WRBs used up by the packet.
930 static u32
be_xmit_enqueue(struct be_adapter
*adapter
, struct be_tx_obj
*txo
,
932 struct be_wrb_params
*wrb_params
)
934 u32 i
, copied
= 0, wrb_cnt
= skb_wrb_cnt(skb
);
935 struct device
*dev
= &adapter
->pdev
->dev
;
936 struct be_queue_info
*txq
= &txo
->q
;
937 bool map_single
= false;
938 u32 head
= txq
->head
;
942 head
= be_tx_get_wrb_hdr(txo
);
944 if (skb
->len
> skb
->data_len
) {
945 len
= skb_headlen(skb
);
947 busaddr
= dma_map_single(dev
, skb
->data
, len
, DMA_TO_DEVICE
);
948 if (dma_mapping_error(dev
, busaddr
))
951 be_tx_setup_wrb_frag(txo
, busaddr
, len
);
955 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
956 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
];
957 len
= skb_frag_size(frag
);
959 busaddr
= skb_frag_dma_map(dev
, frag
, 0, len
, DMA_TO_DEVICE
);
960 if (dma_mapping_error(dev
, busaddr
))
962 be_tx_setup_wrb_frag(txo
, busaddr
, len
);
966 be_tx_setup_wrb_hdr(adapter
, txo
, wrb_params
, skb
, head
);
968 be_tx_stats_update(txo
, skb
);
972 adapter
->drv_stats
.dma_map_errors
++;
973 be_xmit_restore(adapter
, txo
, head
, map_single
, copied
);
977 static inline int qnq_async_evt_rcvd(struct be_adapter
*adapter
)
979 return adapter
->flags
& BE_FLAGS_QNQ_ASYNC_EVT_RCVD
;
982 static struct sk_buff
*be_insert_vlan_in_pkt(struct be_adapter
*adapter
,
989 skb
= skb_share_check(skb
, GFP_ATOMIC
);
993 if (skb_vlan_tag_present(skb
))
994 vlan_tag
= be_get_tx_vlan_tag(adapter
, skb
);
996 if (qnq_async_evt_rcvd(adapter
) && adapter
->pvid
) {
998 vlan_tag
= adapter
->pvid
;
999 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
1000 * skip VLAN insertion
1002 BE_WRB_F_SET(wrb_params
->features
, VLAN_SKIP_HW
, 1);
1006 skb
= vlan_insert_tag_set_proto(skb
, htons(ETH_P_8021Q
),
1013 /* Insert the outer VLAN, if any */
1014 if (adapter
->qnq_vid
) {
1015 vlan_tag
= adapter
->qnq_vid
;
1016 skb
= vlan_insert_tag_set_proto(skb
, htons(ETH_P_8021Q
),
1020 BE_WRB_F_SET(wrb_params
->features
, VLAN_SKIP_HW
, 1);
1026 static bool be_ipv6_exthdr_check(struct sk_buff
*skb
)
1028 struct ethhdr
*eh
= (struct ethhdr
*)skb
->data
;
1029 u16 offset
= ETH_HLEN
;
1031 if (eh
->h_proto
== htons(ETH_P_IPV6
)) {
1032 struct ipv6hdr
*ip6h
= (struct ipv6hdr
*)(skb
->data
+ offset
);
1034 offset
+= sizeof(struct ipv6hdr
);
1035 if (ip6h
->nexthdr
!= NEXTHDR_TCP
&&
1036 ip6h
->nexthdr
!= NEXTHDR_UDP
) {
1037 struct ipv6_opt_hdr
*ehdr
=
1038 (struct ipv6_opt_hdr
*)(skb
->data
+ offset
);
1040 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
1041 if (ehdr
->hdrlen
== 0xff)
1048 static int be_vlan_tag_tx_chk(struct be_adapter
*adapter
, struct sk_buff
*skb
)
1050 return skb_vlan_tag_present(skb
) || adapter
->pvid
|| adapter
->qnq_vid
;
1053 static int be_ipv6_tx_stall_chk(struct be_adapter
*adapter
, struct sk_buff
*skb
)
1055 return BE3_chip(adapter
) && be_ipv6_exthdr_check(skb
);
1058 static struct sk_buff
*be_lancer_xmit_workarounds(struct be_adapter
*adapter
,
1059 struct sk_buff
*skb
,
1060 struct be_wrb_params
1063 struct vlan_ethhdr
*veh
= (struct vlan_ethhdr
*)skb
->data
;
1064 unsigned int eth_hdr_len
;
1067 /* For padded packets, BE HW modifies tot_len field in IP header
1068 * incorrecly when VLAN tag is inserted by HW.
1069 * For padded packets, Lancer computes incorrect checksum.
1071 eth_hdr_len
= ntohs(skb
->protocol
) == ETH_P_8021Q
?
1072 VLAN_ETH_HLEN
: ETH_HLEN
;
1073 if (skb
->len
<= 60 &&
1074 (lancer_chip(adapter
) || skb_vlan_tag_present(skb
)) &&
1076 ip
= (struct iphdr
*)ip_hdr(skb
);
1077 pskb_trim(skb
, eth_hdr_len
+ ntohs(ip
->tot_len
));
1080 /* If vlan tag is already inlined in the packet, skip HW VLAN
1081 * tagging in pvid-tagging mode
1083 if (be_pvid_tagging_enabled(adapter
) &&
1084 veh
->h_vlan_proto
== htons(ETH_P_8021Q
))
1085 BE_WRB_F_SET(wrb_params
->features
, VLAN_SKIP_HW
, 1);
1087 /* HW has a bug wherein it will calculate CSUM for VLAN
1088 * pkts even though it is disabled.
1089 * Manually insert VLAN in pkt.
1091 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
&&
1092 skb_vlan_tag_present(skb
)) {
1093 skb
= be_insert_vlan_in_pkt(adapter
, skb
, wrb_params
);
1098 /* HW may lockup when VLAN HW tagging is requested on
1099 * certain ipv6 packets. Drop such pkts if the HW workaround to
1100 * skip HW tagging is not enabled by FW.
1102 if (unlikely(be_ipv6_tx_stall_chk(adapter
, skb
) &&
1103 (adapter
->pvid
|| adapter
->qnq_vid
) &&
1104 !qnq_async_evt_rcvd(adapter
)))
1107 /* Manual VLAN tag insertion to prevent:
1108 * ASIC lockup when the ASIC inserts VLAN tag into
1109 * certain ipv6 packets. Insert VLAN tags in driver,
1110 * and set event, completion, vlan bits accordingly
1113 if (be_ipv6_tx_stall_chk(adapter
, skb
) &&
1114 be_vlan_tag_tx_chk(adapter
, skb
)) {
1115 skb
= be_insert_vlan_in_pkt(adapter
, skb
, wrb_params
);
1122 dev_kfree_skb_any(skb
);
1127 static struct sk_buff
*be_xmit_workarounds(struct be_adapter
*adapter
,
1128 struct sk_buff
*skb
,
1129 struct be_wrb_params
*wrb_params
)
1133 /* Lancer, SH and BE3 in SRIOV mode have a bug wherein
1134 * packets that are 32b or less may cause a transmit stall
1135 * on that port. The workaround is to pad such packets
1136 * (len <= 32 bytes) to a minimum length of 36b.
1138 if (skb
->len
<= 32) {
1139 if (skb_put_padto(skb
, 36))
1143 if (BEx_chip(adapter
) || lancer_chip(adapter
)) {
1144 skb
= be_lancer_xmit_workarounds(adapter
, skb
, wrb_params
);
1149 /* The stack can send us skbs with length greater than
1150 * what the HW can handle. Trim the extra bytes.
1152 WARN_ON_ONCE(skb
->len
> BE_MAX_GSO_SIZE
);
1153 err
= pskb_trim(skb
, BE_MAX_GSO_SIZE
);
1159 static void be_xmit_flush(struct be_adapter
*adapter
, struct be_tx_obj
*txo
)
1161 struct be_queue_info
*txq
= &txo
->q
;
1162 struct be_eth_hdr_wrb
*hdr
= queue_index_node(txq
, txo
->last_req_hdr
);
1164 /* Mark the last request eventable if it hasn't been marked already */
1165 if (!(hdr
->dw
[2] & cpu_to_le32(TX_HDR_WRB_EVT
)))
1166 hdr
->dw
[2] |= cpu_to_le32(TX_HDR_WRB_EVT
| TX_HDR_WRB_COMPL
);
1168 /* compose a dummy wrb if there are odd set of wrbs to notify */
1169 if (!lancer_chip(adapter
) && (txo
->pend_wrb_cnt
& 1)) {
1170 wrb_fill_dummy(queue_head_node(txq
));
1171 queue_head_inc(txq
);
1172 atomic_inc(&txq
->used
);
1173 txo
->pend_wrb_cnt
++;
1174 hdr
->dw
[2] &= ~cpu_to_le32(TX_HDR_WRB_NUM_MASK
<<
1175 TX_HDR_WRB_NUM_SHIFT
);
1176 hdr
->dw
[2] |= cpu_to_le32((txo
->last_req_wrb_cnt
+ 1) <<
1177 TX_HDR_WRB_NUM_SHIFT
);
1179 be_txq_notify(adapter
, txo
, txo
->pend_wrb_cnt
);
1180 txo
->pend_wrb_cnt
= 0;
1183 /* OS2BMC related */
1185 #define DHCP_CLIENT_PORT 68
1186 #define DHCP_SERVER_PORT 67
1187 #define NET_BIOS_PORT1 137
1188 #define NET_BIOS_PORT2 138
1189 #define DHCPV6_RAS_PORT 547
1191 #define is_mc_allowed_on_bmc(adapter, eh) \
1192 (!is_multicast_filt_enabled(adapter) && \
1193 is_multicast_ether_addr(eh->h_dest) && \
1194 !is_broadcast_ether_addr(eh->h_dest))
1196 #define is_bc_allowed_on_bmc(adapter, eh) \
1197 (!is_broadcast_filt_enabled(adapter) && \
1198 is_broadcast_ether_addr(eh->h_dest))
1200 #define is_arp_allowed_on_bmc(adapter, skb) \
1201 (is_arp(skb) && is_arp_filt_enabled(adapter))
1203 #define is_broadcast_packet(eh, adapter) \
1204 (is_multicast_ether_addr(eh->h_dest) && \
1205 !compare_ether_addr(eh->h_dest, adapter->netdev->broadcast))
1207 #define is_arp(skb) (skb->protocol == htons(ETH_P_ARP))
1209 #define is_arp_filt_enabled(adapter) \
1210 (adapter->bmc_filt_mask & (BMC_FILT_BROADCAST_ARP))
1212 #define is_dhcp_client_filt_enabled(adapter) \
1213 (adapter->bmc_filt_mask & BMC_FILT_BROADCAST_DHCP_CLIENT)
1215 #define is_dhcp_srvr_filt_enabled(adapter) \
1216 (adapter->bmc_filt_mask & BMC_FILT_BROADCAST_DHCP_SERVER)
1218 #define is_nbios_filt_enabled(adapter) \
1219 (adapter->bmc_filt_mask & BMC_FILT_BROADCAST_NET_BIOS)
1221 #define is_ipv6_na_filt_enabled(adapter) \
1222 (adapter->bmc_filt_mask & \
1223 BMC_FILT_MULTICAST_IPV6_NEIGH_ADVER)
1225 #define is_ipv6_ra_filt_enabled(adapter) \
1226 (adapter->bmc_filt_mask & BMC_FILT_MULTICAST_IPV6_RA)
1228 #define is_ipv6_ras_filt_enabled(adapter) \
1229 (adapter->bmc_filt_mask & BMC_FILT_MULTICAST_IPV6_RAS)
1231 #define is_broadcast_filt_enabled(adapter) \
1232 (adapter->bmc_filt_mask & BMC_FILT_BROADCAST)
1234 #define is_multicast_filt_enabled(adapter) \
1235 (adapter->bmc_filt_mask & BMC_FILT_MULTICAST)
1237 static bool be_send_pkt_to_bmc(struct be_adapter
*adapter
,
1238 struct sk_buff
**skb
)
1240 struct ethhdr
*eh
= (struct ethhdr
*)(*skb
)->data
;
1241 bool os2bmc
= false;
1243 if (!be_is_os2bmc_enabled(adapter
))
1246 if (!is_multicast_ether_addr(eh
->h_dest
))
1249 if (is_mc_allowed_on_bmc(adapter
, eh
) ||
1250 is_bc_allowed_on_bmc(adapter
, eh
) ||
1251 is_arp_allowed_on_bmc(adapter
, (*skb
))) {
1256 if ((*skb
)->protocol
== htons(ETH_P_IPV6
)) {
1257 struct ipv6hdr
*hdr
= ipv6_hdr((*skb
));
1258 u8 nexthdr
= hdr
->nexthdr
;
1260 if (nexthdr
== IPPROTO_ICMPV6
) {
1261 struct icmp6hdr
*icmp6
= icmp6_hdr((*skb
));
1263 switch (icmp6
->icmp6_type
) {
1264 case NDISC_ROUTER_ADVERTISEMENT
:
1265 os2bmc
= is_ipv6_ra_filt_enabled(adapter
);
1267 case NDISC_NEIGHBOUR_ADVERTISEMENT
:
1268 os2bmc
= is_ipv6_na_filt_enabled(adapter
);
1276 if (is_udp_pkt((*skb
))) {
1277 struct udphdr
*udp
= udp_hdr((*skb
));
1279 switch (ntohs(udp
->dest
)) {
1280 case DHCP_CLIENT_PORT
:
1281 os2bmc
= is_dhcp_client_filt_enabled(adapter
);
1283 case DHCP_SERVER_PORT
:
1284 os2bmc
= is_dhcp_srvr_filt_enabled(adapter
);
1286 case NET_BIOS_PORT1
:
1287 case NET_BIOS_PORT2
:
1288 os2bmc
= is_nbios_filt_enabled(adapter
);
1290 case DHCPV6_RAS_PORT
:
1291 os2bmc
= is_ipv6_ras_filt_enabled(adapter
);
1298 /* For packets over a vlan, which are destined
1299 * to BMC, asic expects the vlan to be inline in the packet.
1302 *skb
= be_insert_vlan_in_pkt(adapter
, *skb
, NULL
);
1307 static netdev_tx_t
be_xmit(struct sk_buff
*skb
, struct net_device
*netdev
)
1309 struct be_adapter
*adapter
= netdev_priv(netdev
);
1310 u16 q_idx
= skb_get_queue_mapping(skb
);
1311 struct be_tx_obj
*txo
= &adapter
->tx_obj
[q_idx
];
1312 struct be_wrb_params wrb_params
= { 0 };
1313 bool flush
= !skb
->xmit_more
;
1316 skb
= be_xmit_workarounds(adapter
, skb
, &wrb_params
);
1320 be_get_wrb_params_from_skb(adapter
, skb
, &wrb_params
);
1322 wrb_cnt
= be_xmit_enqueue(adapter
, txo
, skb
, &wrb_params
);
1323 if (unlikely(!wrb_cnt
)) {
1324 dev_kfree_skb_any(skb
);
1328 /* if os2bmc is enabled and if the pkt is destined to bmc,
1329 * enqueue the pkt a 2nd time with mgmt bit set.
1331 if (be_send_pkt_to_bmc(adapter
, &skb
)) {
1332 BE_WRB_F_SET(wrb_params
.features
, OS2BMC
, 1);
1333 wrb_cnt
= be_xmit_enqueue(adapter
, txo
, skb
, &wrb_params
);
1334 if (unlikely(!wrb_cnt
))
1340 if (be_is_txq_full(txo
)) {
1341 netif_stop_subqueue(netdev
, q_idx
);
1342 tx_stats(txo
)->tx_stops
++;
1345 if (flush
|| __netif_subqueue_stopped(netdev
, q_idx
))
1346 be_xmit_flush(adapter
, txo
);
1348 return NETDEV_TX_OK
;
1350 tx_stats(txo
)->tx_drv_drops
++;
1351 /* Flush the already enqueued tx requests */
1352 if (flush
&& txo
->pend_wrb_cnt
)
1353 be_xmit_flush(adapter
, txo
);
1355 return NETDEV_TX_OK
;
1358 static int be_change_mtu(struct net_device
*netdev
, int new_mtu
)
1360 struct be_adapter
*adapter
= netdev_priv(netdev
);
1361 struct device
*dev
= &adapter
->pdev
->dev
;
1363 if (new_mtu
< BE_MIN_MTU
|| new_mtu
> BE_MAX_MTU
) {
1364 dev_info(dev
, "MTU must be between %d and %d bytes\n",
1365 BE_MIN_MTU
, BE_MAX_MTU
);
1369 dev_info(dev
, "MTU changed from %d to %d bytes\n",
1370 netdev
->mtu
, new_mtu
);
1371 netdev
->mtu
= new_mtu
;
1375 static inline bool be_in_all_promisc(struct be_adapter
*adapter
)
1377 return (adapter
->if_flags
& BE_IF_FLAGS_ALL_PROMISCUOUS
) ==
1378 BE_IF_FLAGS_ALL_PROMISCUOUS
;
1381 static int be_set_vlan_promisc(struct be_adapter
*adapter
)
1383 struct device
*dev
= &adapter
->pdev
->dev
;
1386 if (adapter
->if_flags
& BE_IF_FLAGS_VLAN_PROMISCUOUS
)
1389 status
= be_cmd_rx_filter(adapter
, BE_IF_FLAGS_VLAN_PROMISCUOUS
, ON
);
1391 dev_info(dev
, "Enabled VLAN promiscuous mode\n");
1392 adapter
->if_flags
|= BE_IF_FLAGS_VLAN_PROMISCUOUS
;
1394 dev_err(dev
, "Failed to enable VLAN promiscuous mode\n");
1399 static int be_clear_vlan_promisc(struct be_adapter
*adapter
)
1401 struct device
*dev
= &adapter
->pdev
->dev
;
1404 status
= be_cmd_rx_filter(adapter
, BE_IF_FLAGS_VLAN_PROMISCUOUS
, OFF
);
1406 dev_info(dev
, "Disabling VLAN promiscuous mode\n");
1407 adapter
->if_flags
&= ~BE_IF_FLAGS_VLAN_PROMISCUOUS
;
1413 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1414 * If the user configures more, place BE in vlan promiscuous mode.
1416 static int be_vid_config(struct be_adapter
*adapter
)
1418 struct device
*dev
= &adapter
->pdev
->dev
;
1419 u16 vids
[BE_NUM_VLANS_SUPPORTED
];
1423 /* No need to further configure vids if in promiscuous mode */
1424 if (be_in_all_promisc(adapter
))
1427 if (adapter
->vlans_added
> be_max_vlans(adapter
))
1428 return be_set_vlan_promisc(adapter
);
1430 /* Construct VLAN Table to give to HW */
1431 for_each_set_bit(i
, adapter
->vids
, VLAN_N_VID
)
1432 vids
[num
++] = cpu_to_le16(i
);
1434 status
= be_cmd_vlan_config(adapter
, adapter
->if_handle
, vids
, num
, 0);
1436 dev_err(dev
, "Setting HW VLAN filtering failed\n");
1437 /* Set to VLAN promisc mode as setting VLAN filter failed */
1438 if (addl_status(status
) == MCC_ADDL_STATUS_INSUFFICIENT_VLANS
||
1439 addl_status(status
) ==
1440 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES
)
1441 return be_set_vlan_promisc(adapter
);
1442 } else if (adapter
->if_flags
& BE_IF_FLAGS_VLAN_PROMISCUOUS
) {
1443 status
= be_clear_vlan_promisc(adapter
);
1448 static int be_vlan_add_vid(struct net_device
*netdev
, __be16 proto
, u16 vid
)
1450 struct be_adapter
*adapter
= netdev_priv(netdev
);
1453 /* Packets with VID 0 are always received by Lancer by default */
1454 if (lancer_chip(adapter
) && vid
== 0)
1457 if (test_bit(vid
, adapter
->vids
))
1460 set_bit(vid
, adapter
->vids
);
1461 adapter
->vlans_added
++;
1463 status
= be_vid_config(adapter
);
1465 adapter
->vlans_added
--;
1466 clear_bit(vid
, adapter
->vids
);
1472 static int be_vlan_rem_vid(struct net_device
*netdev
, __be16 proto
, u16 vid
)
1474 struct be_adapter
*adapter
= netdev_priv(netdev
);
1476 /* Packets with VID 0 are always received by Lancer by default */
1477 if (lancer_chip(adapter
) && vid
== 0)
1480 if (!test_bit(vid
, adapter
->vids
))
1483 clear_bit(vid
, adapter
->vids
);
1484 adapter
->vlans_added
--;
1486 return be_vid_config(adapter
);
1489 static void be_clear_all_promisc(struct be_adapter
*adapter
)
1491 be_cmd_rx_filter(adapter
, BE_IF_FLAGS_ALL_PROMISCUOUS
, OFF
);
1492 adapter
->if_flags
&= ~BE_IF_FLAGS_ALL_PROMISCUOUS
;
1495 static void be_set_all_promisc(struct be_adapter
*adapter
)
1497 be_cmd_rx_filter(adapter
, BE_IF_FLAGS_ALL_PROMISCUOUS
, ON
);
1498 adapter
->if_flags
|= BE_IF_FLAGS_ALL_PROMISCUOUS
;
1501 static void be_set_mc_promisc(struct be_adapter
*adapter
)
1505 if (adapter
->if_flags
& BE_IF_FLAGS_MCAST_PROMISCUOUS
)
1508 status
= be_cmd_rx_filter(adapter
, BE_IF_FLAGS_MCAST_PROMISCUOUS
, ON
);
1510 adapter
->if_flags
|= BE_IF_FLAGS_MCAST_PROMISCUOUS
;
1513 static void be_set_mc_list(struct be_adapter
*adapter
)
1517 status
= be_cmd_rx_filter(adapter
, BE_IF_FLAGS_MULTICAST
, ON
);
1519 adapter
->if_flags
&= ~BE_IF_FLAGS_MCAST_PROMISCUOUS
;
1521 be_set_mc_promisc(adapter
);
1524 static void be_set_uc_list(struct be_adapter
*adapter
)
1526 struct netdev_hw_addr
*ha
;
1527 int i
= 1; /* First slot is claimed by the Primary MAC */
1529 for (; adapter
->uc_macs
> 0; adapter
->uc_macs
--, i
++)
1530 be_cmd_pmac_del(adapter
, adapter
->if_handle
,
1531 adapter
->pmac_id
[i
], 0);
1533 if (netdev_uc_count(adapter
->netdev
) > be_max_uc(adapter
)) {
1534 be_set_all_promisc(adapter
);
1538 netdev_for_each_uc_addr(ha
, adapter
->netdev
) {
1539 adapter
->uc_macs
++; /* First slot is for Primary MAC */
1540 be_cmd_pmac_add(adapter
, (u8
*)ha
->addr
, adapter
->if_handle
,
1541 &adapter
->pmac_id
[adapter
->uc_macs
], 0);
1545 static void be_clear_uc_list(struct be_adapter
*adapter
)
1549 for (i
= 1; i
< (adapter
->uc_macs
+ 1); i
++)
1550 be_cmd_pmac_del(adapter
, adapter
->if_handle
,
1551 adapter
->pmac_id
[i
], 0);
1552 adapter
->uc_macs
= 0;
1555 static void be_set_rx_mode(struct net_device
*netdev
)
1557 struct be_adapter
*adapter
= netdev_priv(netdev
);
1559 if (netdev
->flags
& IFF_PROMISC
) {
1560 be_set_all_promisc(adapter
);
1564 /* Interface was previously in promiscuous mode; disable it */
1565 if (be_in_all_promisc(adapter
)) {
1566 be_clear_all_promisc(adapter
);
1567 if (adapter
->vlans_added
)
1568 be_vid_config(adapter
);
1571 /* Enable multicast promisc if num configured exceeds what we support */
1572 if (netdev
->flags
& IFF_ALLMULTI
||
1573 netdev_mc_count(netdev
) > be_max_mc(adapter
)) {
1574 be_set_mc_promisc(adapter
);
1578 if (netdev_uc_count(netdev
) != adapter
->uc_macs
)
1579 be_set_uc_list(adapter
);
1581 be_set_mc_list(adapter
);
1584 static int be_set_vf_mac(struct net_device
*netdev
, int vf
, u8
*mac
)
1586 struct be_adapter
*adapter
= netdev_priv(netdev
);
1587 struct be_vf_cfg
*vf_cfg
= &adapter
->vf_cfg
[vf
];
1590 if (!sriov_enabled(adapter
))
1593 if (!is_valid_ether_addr(mac
) || vf
>= adapter
->num_vfs
)
1596 /* Proceed further only if user provided MAC is different
1599 if (ether_addr_equal(mac
, vf_cfg
->mac_addr
))
1602 if (BEx_chip(adapter
)) {
1603 be_cmd_pmac_del(adapter
, vf_cfg
->if_handle
, vf_cfg
->pmac_id
,
1606 status
= be_cmd_pmac_add(adapter
, mac
, vf_cfg
->if_handle
,
1607 &vf_cfg
->pmac_id
, vf
+ 1);
1609 status
= be_cmd_set_mac(adapter
, mac
, vf_cfg
->if_handle
,
1614 dev_err(&adapter
->pdev
->dev
, "MAC %pM set on VF %d Failed: %#x",
1616 return be_cmd_status(status
);
1619 ether_addr_copy(vf_cfg
->mac_addr
, mac
);
1624 static int be_get_vf_config(struct net_device
*netdev
, int vf
,
1625 struct ifla_vf_info
*vi
)
1627 struct be_adapter
*adapter
= netdev_priv(netdev
);
1628 struct be_vf_cfg
*vf_cfg
= &adapter
->vf_cfg
[vf
];
1630 if (!sriov_enabled(adapter
))
1633 if (vf
>= adapter
->num_vfs
)
1637 vi
->max_tx_rate
= vf_cfg
->tx_rate
;
1638 vi
->min_tx_rate
= 0;
1639 vi
->vlan
= vf_cfg
->vlan_tag
& VLAN_VID_MASK
;
1640 vi
->qos
= vf_cfg
->vlan_tag
>> VLAN_PRIO_SHIFT
;
1641 memcpy(&vi
->mac
, vf_cfg
->mac_addr
, ETH_ALEN
);
1642 vi
->linkstate
= adapter
->vf_cfg
[vf
].plink_tracking
;
1643 vi
->spoofchk
= adapter
->vf_cfg
[vf
].spoofchk
;
1648 static int be_set_vf_tvt(struct be_adapter
*adapter
, int vf
, u16 vlan
)
1650 struct be_vf_cfg
*vf_cfg
= &adapter
->vf_cfg
[vf
];
1651 u16 vids
[BE_NUM_VLANS_SUPPORTED
];
1652 int vf_if_id
= vf_cfg
->if_handle
;
1655 /* Enable Transparent VLAN Tagging */
1656 status
= be_cmd_set_hsw_config(adapter
, vlan
, vf
+ 1, vf_if_id
, 0, 0);
1660 /* Clear pre-programmed VLAN filters on VF if any, if TVT is enabled */
1662 status
= be_cmd_vlan_config(adapter
, vf_if_id
, vids
, 1, vf
+ 1);
1664 dev_info(&adapter
->pdev
->dev
,
1665 "Cleared guest VLANs on VF%d", vf
);
1667 /* After TVT is enabled, disallow VFs to program VLAN filters */
1668 if (vf_cfg
->privileges
& BE_PRIV_FILTMGMT
) {
1669 status
= be_cmd_set_fn_privileges(adapter
, vf_cfg
->privileges
&
1670 ~BE_PRIV_FILTMGMT
, vf
+ 1);
1672 vf_cfg
->privileges
&= ~BE_PRIV_FILTMGMT
;
1677 static int be_clear_vf_tvt(struct be_adapter
*adapter
, int vf
)
1679 struct be_vf_cfg
*vf_cfg
= &adapter
->vf_cfg
[vf
];
1680 struct device
*dev
= &adapter
->pdev
->dev
;
1683 /* Reset Transparent VLAN Tagging. */
1684 status
= be_cmd_set_hsw_config(adapter
, BE_RESET_VLAN_TAG_ID
, vf
+ 1,
1685 vf_cfg
->if_handle
, 0, 0);
1689 /* Allow VFs to program VLAN filtering */
1690 if (!(vf_cfg
->privileges
& BE_PRIV_FILTMGMT
)) {
1691 status
= be_cmd_set_fn_privileges(adapter
, vf_cfg
->privileges
|
1692 BE_PRIV_FILTMGMT
, vf
+ 1);
1694 vf_cfg
->privileges
|= BE_PRIV_FILTMGMT
;
1695 dev_info(dev
, "VF%d: FILTMGMT priv enabled", vf
);
1700 "Disable/re-enable i/f in VM to clear Transparent VLAN tag");
1704 static int be_set_vf_vlan(struct net_device
*netdev
, int vf
, u16 vlan
, u8 qos
)
1706 struct be_adapter
*adapter
= netdev_priv(netdev
);
1707 struct be_vf_cfg
*vf_cfg
= &adapter
->vf_cfg
[vf
];
1710 if (!sriov_enabled(adapter
))
1713 if (vf
>= adapter
->num_vfs
|| vlan
> 4095 || qos
> 7)
1717 vlan
|= qos
<< VLAN_PRIO_SHIFT
;
1718 status
= be_set_vf_tvt(adapter
, vf
, vlan
);
1720 status
= be_clear_vf_tvt(adapter
, vf
);
1724 dev_err(&adapter
->pdev
->dev
,
1725 "VLAN %d config on VF %d failed : %#x\n", vlan
, vf
,
1727 return be_cmd_status(status
);
1730 vf_cfg
->vlan_tag
= vlan
;
1734 static int be_set_vf_tx_rate(struct net_device
*netdev
, int vf
,
1735 int min_tx_rate
, int max_tx_rate
)
1737 struct be_adapter
*adapter
= netdev_priv(netdev
);
1738 struct device
*dev
= &adapter
->pdev
->dev
;
1739 int percent_rate
, status
= 0;
1743 if (!sriov_enabled(adapter
))
1746 if (vf
>= adapter
->num_vfs
)
1755 status
= be_cmd_link_status_query(adapter
, &link_speed
,
1761 dev_err(dev
, "TX-rate setting not allowed when link is down\n");
1766 if (max_tx_rate
< 100 || max_tx_rate
> link_speed
) {
1767 dev_err(dev
, "TX-rate must be between 100 and %d Mbps\n",
1773 /* On Skyhawk the QOS setting must be done only as a % value */
1774 percent_rate
= link_speed
/ 100;
1775 if (skyhawk_chip(adapter
) && (max_tx_rate
% percent_rate
)) {
1776 dev_err(dev
, "TX-rate must be a multiple of %d Mbps\n",
1783 status
= be_cmd_config_qos(adapter
, max_tx_rate
, link_speed
, vf
+ 1);
1787 adapter
->vf_cfg
[vf
].tx_rate
= max_tx_rate
;
1791 dev_err(dev
, "TX-rate setting of %dMbps on VF%d failed\n",
1793 return be_cmd_status(status
);
1796 static int be_set_vf_link_state(struct net_device
*netdev
, int vf
,
1799 struct be_adapter
*adapter
= netdev_priv(netdev
);
1802 if (!sriov_enabled(adapter
))
1805 if (vf
>= adapter
->num_vfs
)
1808 status
= be_cmd_set_logical_link_config(adapter
, link_state
, vf
+1);
1810 dev_err(&adapter
->pdev
->dev
,
1811 "Link state change on VF %d failed: %#x\n", vf
, status
);
1812 return be_cmd_status(status
);
1815 adapter
->vf_cfg
[vf
].plink_tracking
= link_state
;
1820 static int be_set_vf_spoofchk(struct net_device
*netdev
, int vf
, bool enable
)
1822 struct be_adapter
*adapter
= netdev_priv(netdev
);
1823 struct be_vf_cfg
*vf_cfg
= &adapter
->vf_cfg
[vf
];
1827 if (!sriov_enabled(adapter
))
1830 if (vf
>= adapter
->num_vfs
)
1833 if (BEx_chip(adapter
))
1836 if (enable
== vf_cfg
->spoofchk
)
1839 spoofchk
= enable
? ENABLE_MAC_SPOOFCHK
: DISABLE_MAC_SPOOFCHK
;
1841 status
= be_cmd_set_hsw_config(adapter
, 0, vf
+ 1, vf_cfg
->if_handle
,
1844 dev_err(&adapter
->pdev
->dev
,
1845 "Spoofchk change on VF %d failed: %#x\n", vf
, status
);
1846 return be_cmd_status(status
);
1849 vf_cfg
->spoofchk
= enable
;
1853 static void be_aic_update(struct be_aic_obj
*aic
, u64 rx_pkts
, u64 tx_pkts
,
1856 aic
->rx_pkts_prev
= rx_pkts
;
1857 aic
->tx_reqs_prev
= tx_pkts
;
1861 static int be_get_new_eqd(struct be_eq_obj
*eqo
)
1863 struct be_adapter
*adapter
= eqo
->adapter
;
1865 struct be_aic_obj
*aic
;
1866 struct be_rx_obj
*rxo
;
1867 struct be_tx_obj
*txo
;
1868 u64 rx_pkts
= 0, tx_pkts
= 0;
1873 aic
= &adapter
->aic_obj
[eqo
->idx
];
1881 for_all_rx_queues_on_eq(adapter
, eqo
, rxo
, i
) {
1883 start
= u64_stats_fetch_begin_irq(&rxo
->stats
.sync
);
1884 rx_pkts
+= rxo
->stats
.rx_pkts
;
1885 } while (u64_stats_fetch_retry_irq(&rxo
->stats
.sync
, start
));
1888 for_all_tx_queues_on_eq(adapter
, eqo
, txo
, i
) {
1890 start
= u64_stats_fetch_begin_irq(&txo
->stats
.sync
);
1891 tx_pkts
+= txo
->stats
.tx_reqs
;
1892 } while (u64_stats_fetch_retry_irq(&txo
->stats
.sync
, start
));
1895 /* Skip, if wrapped around or first calculation */
1897 if (!aic
->jiffies
|| time_before(now
, aic
->jiffies
) ||
1898 rx_pkts
< aic
->rx_pkts_prev
||
1899 tx_pkts
< aic
->tx_reqs_prev
) {
1900 be_aic_update(aic
, rx_pkts
, tx_pkts
, now
);
1901 return aic
->prev_eqd
;
1904 delta
= jiffies_to_msecs(now
- aic
->jiffies
);
1906 return aic
->prev_eqd
;
1908 pps
= (((u32
)(rx_pkts
- aic
->rx_pkts_prev
) * 1000) / delta
) +
1909 (((u32
)(tx_pkts
- aic
->tx_reqs_prev
) * 1000) / delta
);
1910 eqd
= (pps
/ 15000) << 2;
1914 eqd
= min_t(u32
, eqd
, aic
->max_eqd
);
1915 eqd
= max_t(u32
, eqd
, aic
->min_eqd
);
1917 be_aic_update(aic
, rx_pkts
, tx_pkts
, now
);
1922 /* For Skyhawk-R only */
1923 static u32
be_get_eq_delay_mult_enc(struct be_eq_obj
*eqo
)
1925 struct be_adapter
*adapter
= eqo
->adapter
;
1926 struct be_aic_obj
*aic
= &adapter
->aic_obj
[eqo
->idx
];
1927 ulong now
= jiffies
;
1934 if (jiffies_to_msecs(now
- aic
->jiffies
) < 1)
1935 eqd
= aic
->prev_eqd
;
1937 eqd
= be_get_new_eqd(eqo
);
1940 mult_enc
= R2I_DLY_ENC_1
;
1942 mult_enc
= R2I_DLY_ENC_2
;
1944 mult_enc
= R2I_DLY_ENC_3
;
1946 mult_enc
= R2I_DLY_ENC_0
;
1948 aic
->prev_eqd
= eqd
;
1953 void be_eqd_update(struct be_adapter
*adapter
, bool force_update
)
1955 struct be_set_eqd set_eqd
[MAX_EVT_QS
];
1956 struct be_aic_obj
*aic
;
1957 struct be_eq_obj
*eqo
;
1958 int i
, num
= 0, eqd
;
1960 for_all_evt_queues(adapter
, eqo
, i
) {
1961 aic
= &adapter
->aic_obj
[eqo
->idx
];
1962 eqd
= be_get_new_eqd(eqo
);
1963 if (force_update
|| eqd
!= aic
->prev_eqd
) {
1964 set_eqd
[num
].delay_multiplier
= (eqd
* 65)/100;
1965 set_eqd
[num
].eq_id
= eqo
->q
.id
;
1966 aic
->prev_eqd
= eqd
;
1972 be_cmd_modify_eqd(adapter
, set_eqd
, num
);
1975 static void be_rx_stats_update(struct be_rx_obj
*rxo
,
1976 struct be_rx_compl_info
*rxcp
)
1978 struct be_rx_stats
*stats
= rx_stats(rxo
);
1980 u64_stats_update_begin(&stats
->sync
);
1982 stats
->rx_bytes
+= rxcp
->pkt_size
;
1985 stats
->rx_vxlan_offload_pkts
++;
1986 if (rxcp
->pkt_type
== BE_MULTICAST_PACKET
)
1987 stats
->rx_mcast_pkts
++;
1989 stats
->rx_compl_err
++;
1990 u64_stats_update_end(&stats
->sync
);
1993 static inline bool csum_passed(struct be_rx_compl_info
*rxcp
)
1995 /* L4 checksum is not reliable for non TCP/UDP packets.
1996 * Also ignore ipcksm for ipv6 pkts
1998 return (rxcp
->tcpf
|| rxcp
->udpf
) && rxcp
->l4_csum
&&
1999 (rxcp
->ip_csum
|| rxcp
->ipv6
) && !rxcp
->err
;
2002 static struct be_rx_page_info
*get_rx_page_info(struct be_rx_obj
*rxo
)
2004 struct be_adapter
*adapter
= rxo
->adapter
;
2005 struct be_rx_page_info
*rx_page_info
;
2006 struct be_queue_info
*rxq
= &rxo
->q
;
2007 u32 frag_idx
= rxq
->tail
;
2009 rx_page_info
= &rxo
->page_info_tbl
[frag_idx
];
2010 BUG_ON(!rx_page_info
->page
);
2012 if (rx_page_info
->last_frag
) {
2013 dma_unmap_page(&adapter
->pdev
->dev
,
2014 dma_unmap_addr(rx_page_info
, bus
),
2015 adapter
->big_page_size
, DMA_FROM_DEVICE
);
2016 rx_page_info
->last_frag
= false;
2018 dma_sync_single_for_cpu(&adapter
->pdev
->dev
,
2019 dma_unmap_addr(rx_page_info
, bus
),
2020 rx_frag_size
, DMA_FROM_DEVICE
);
2023 queue_tail_inc(rxq
);
2024 atomic_dec(&rxq
->used
);
2025 return rx_page_info
;
2028 /* Throwaway the data in the Rx completion */
2029 static void be_rx_compl_discard(struct be_rx_obj
*rxo
,
2030 struct be_rx_compl_info
*rxcp
)
2032 struct be_rx_page_info
*page_info
;
2033 u16 i
, num_rcvd
= rxcp
->num_rcvd
;
2035 for (i
= 0; i
< num_rcvd
; i
++) {
2036 page_info
= get_rx_page_info(rxo
);
2037 put_page(page_info
->page
);
2038 memset(page_info
, 0, sizeof(*page_info
));
2043 * skb_fill_rx_data forms a complete skb for an ether frame
2044 * indicated by rxcp.
2046 static void skb_fill_rx_data(struct be_rx_obj
*rxo
, struct sk_buff
*skb
,
2047 struct be_rx_compl_info
*rxcp
)
2049 struct be_rx_page_info
*page_info
;
2051 u16 hdr_len
, curr_frag_len
, remaining
;
2054 page_info
= get_rx_page_info(rxo
);
2055 start
= page_address(page_info
->page
) + page_info
->page_offset
;
2058 /* Copy data in the first descriptor of this completion */
2059 curr_frag_len
= min(rxcp
->pkt_size
, rx_frag_size
);
2061 skb
->len
= curr_frag_len
;
2062 if (curr_frag_len
<= BE_HDR_LEN
) { /* tiny packet */
2063 memcpy(skb
->data
, start
, curr_frag_len
);
2064 /* Complete packet has now been moved to data */
2065 put_page(page_info
->page
);
2067 skb
->tail
+= curr_frag_len
;
2070 memcpy(skb
->data
, start
, hdr_len
);
2071 skb_shinfo(skb
)->nr_frags
= 1;
2072 skb_frag_set_page(skb
, 0, page_info
->page
);
2073 skb_shinfo(skb
)->frags
[0].page_offset
=
2074 page_info
->page_offset
+ hdr_len
;
2075 skb_frag_size_set(&skb_shinfo(skb
)->frags
[0],
2076 curr_frag_len
- hdr_len
);
2077 skb
->data_len
= curr_frag_len
- hdr_len
;
2078 skb
->truesize
+= rx_frag_size
;
2079 skb
->tail
+= hdr_len
;
2081 page_info
->page
= NULL
;
2083 if (rxcp
->pkt_size
<= rx_frag_size
) {
2084 BUG_ON(rxcp
->num_rcvd
!= 1);
2088 /* More frags present for this completion */
2089 remaining
= rxcp
->pkt_size
- curr_frag_len
;
2090 for (i
= 1, j
= 0; i
< rxcp
->num_rcvd
; i
++) {
2091 page_info
= get_rx_page_info(rxo
);
2092 curr_frag_len
= min(remaining
, rx_frag_size
);
2094 /* Coalesce all frags from the same physical page in one slot */
2095 if (page_info
->page_offset
== 0) {
2098 skb_frag_set_page(skb
, j
, page_info
->page
);
2099 skb_shinfo(skb
)->frags
[j
].page_offset
=
2100 page_info
->page_offset
;
2101 skb_frag_size_set(&skb_shinfo(skb
)->frags
[j
], 0);
2102 skb_shinfo(skb
)->nr_frags
++;
2104 put_page(page_info
->page
);
2107 skb_frag_size_add(&skb_shinfo(skb
)->frags
[j
], curr_frag_len
);
2108 skb
->len
+= curr_frag_len
;
2109 skb
->data_len
+= curr_frag_len
;
2110 skb
->truesize
+= rx_frag_size
;
2111 remaining
-= curr_frag_len
;
2112 page_info
->page
= NULL
;
2114 BUG_ON(j
> MAX_SKB_FRAGS
);
2117 /* Process the RX completion indicated by rxcp when GRO is disabled */
2118 static void be_rx_compl_process(struct be_rx_obj
*rxo
, struct napi_struct
*napi
,
2119 struct be_rx_compl_info
*rxcp
)
2121 struct be_adapter
*adapter
= rxo
->adapter
;
2122 struct net_device
*netdev
= adapter
->netdev
;
2123 struct sk_buff
*skb
;
2125 skb
= netdev_alloc_skb_ip_align(netdev
, BE_RX_SKB_ALLOC_SIZE
);
2126 if (unlikely(!skb
)) {
2127 rx_stats(rxo
)->rx_drops_no_skbs
++;
2128 be_rx_compl_discard(rxo
, rxcp
);
2132 skb_fill_rx_data(rxo
, skb
, rxcp
);
2134 if (likely((netdev
->features
& NETIF_F_RXCSUM
) && csum_passed(rxcp
)))
2135 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2137 skb_checksum_none_assert(skb
);
2139 skb
->protocol
= eth_type_trans(skb
, netdev
);
2140 skb_record_rx_queue(skb
, rxo
- &adapter
->rx_obj
[0]);
2141 if (netdev
->features
& NETIF_F_RXHASH
)
2142 skb_set_hash(skb
, rxcp
->rss_hash
, PKT_HASH_TYPE_L3
);
2144 skb
->csum_level
= rxcp
->tunneled
;
2145 skb_mark_napi_id(skb
, napi
);
2148 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), rxcp
->vlan_tag
);
2150 netif_receive_skb(skb
);
2153 /* Process the RX completion indicated by rxcp when GRO is enabled */
2154 static void be_rx_compl_process_gro(struct be_rx_obj
*rxo
,
2155 struct napi_struct
*napi
,
2156 struct be_rx_compl_info
*rxcp
)
2158 struct be_adapter
*adapter
= rxo
->adapter
;
2159 struct be_rx_page_info
*page_info
;
2160 struct sk_buff
*skb
= NULL
;
2161 u16 remaining
, curr_frag_len
;
2164 skb
= napi_get_frags(napi
);
2166 be_rx_compl_discard(rxo
, rxcp
);
2170 remaining
= rxcp
->pkt_size
;
2171 for (i
= 0, j
= -1; i
< rxcp
->num_rcvd
; i
++) {
2172 page_info
= get_rx_page_info(rxo
);
2174 curr_frag_len
= min(remaining
, rx_frag_size
);
2176 /* Coalesce all frags from the same physical page in one slot */
2177 if (i
== 0 || page_info
->page_offset
== 0) {
2178 /* First frag or Fresh page */
2180 skb_frag_set_page(skb
, j
, page_info
->page
);
2181 skb_shinfo(skb
)->frags
[j
].page_offset
=
2182 page_info
->page_offset
;
2183 skb_frag_size_set(&skb_shinfo(skb
)->frags
[j
], 0);
2185 put_page(page_info
->page
);
2187 skb_frag_size_add(&skb_shinfo(skb
)->frags
[j
], curr_frag_len
);
2188 skb
->truesize
+= rx_frag_size
;
2189 remaining
-= curr_frag_len
;
2190 memset(page_info
, 0, sizeof(*page_info
));
2192 BUG_ON(j
> MAX_SKB_FRAGS
);
2194 skb_shinfo(skb
)->nr_frags
= j
+ 1;
2195 skb
->len
= rxcp
->pkt_size
;
2196 skb
->data_len
= rxcp
->pkt_size
;
2197 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2198 skb_record_rx_queue(skb
, rxo
- &adapter
->rx_obj
[0]);
2199 if (adapter
->netdev
->features
& NETIF_F_RXHASH
)
2200 skb_set_hash(skb
, rxcp
->rss_hash
, PKT_HASH_TYPE_L3
);
2202 skb
->csum_level
= rxcp
->tunneled
;
2205 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), rxcp
->vlan_tag
);
2207 napi_gro_frags(napi
);
2210 static void be_parse_rx_compl_v1(struct be_eth_rx_compl
*compl,
2211 struct be_rx_compl_info
*rxcp
)
2213 rxcp
->pkt_size
= GET_RX_COMPL_V1_BITS(pktsize
, compl);
2214 rxcp
->vlanf
= GET_RX_COMPL_V1_BITS(vtp
, compl);
2215 rxcp
->err
= GET_RX_COMPL_V1_BITS(err
, compl);
2216 rxcp
->tcpf
= GET_RX_COMPL_V1_BITS(tcpf
, compl);
2217 rxcp
->udpf
= GET_RX_COMPL_V1_BITS(udpf
, compl);
2218 rxcp
->ip_csum
= GET_RX_COMPL_V1_BITS(ipcksm
, compl);
2219 rxcp
->l4_csum
= GET_RX_COMPL_V1_BITS(l4_cksm
, compl);
2220 rxcp
->ipv6
= GET_RX_COMPL_V1_BITS(ip_version
, compl);
2221 rxcp
->num_rcvd
= GET_RX_COMPL_V1_BITS(numfrags
, compl);
2222 rxcp
->pkt_type
= GET_RX_COMPL_V1_BITS(cast_enc
, compl);
2223 rxcp
->rss_hash
= GET_RX_COMPL_V1_BITS(rsshash
, compl);
2225 rxcp
->qnq
= GET_RX_COMPL_V1_BITS(qnq
, compl);
2226 rxcp
->vlan_tag
= GET_RX_COMPL_V1_BITS(vlan_tag
, compl);
2228 rxcp
->port
= GET_RX_COMPL_V1_BITS(port
, compl);
2230 GET_RX_COMPL_V1_BITS(tunneled
, compl);
2233 static void be_parse_rx_compl_v0(struct be_eth_rx_compl
*compl,
2234 struct be_rx_compl_info
*rxcp
)
2236 rxcp
->pkt_size
= GET_RX_COMPL_V0_BITS(pktsize
, compl);
2237 rxcp
->vlanf
= GET_RX_COMPL_V0_BITS(vtp
, compl);
2238 rxcp
->err
= GET_RX_COMPL_V0_BITS(err
, compl);
2239 rxcp
->tcpf
= GET_RX_COMPL_V0_BITS(tcpf
, compl);
2240 rxcp
->udpf
= GET_RX_COMPL_V0_BITS(udpf
, compl);
2241 rxcp
->ip_csum
= GET_RX_COMPL_V0_BITS(ipcksm
, compl);
2242 rxcp
->l4_csum
= GET_RX_COMPL_V0_BITS(l4_cksm
, compl);
2243 rxcp
->ipv6
= GET_RX_COMPL_V0_BITS(ip_version
, compl);
2244 rxcp
->num_rcvd
= GET_RX_COMPL_V0_BITS(numfrags
, compl);
2245 rxcp
->pkt_type
= GET_RX_COMPL_V0_BITS(cast_enc
, compl);
2246 rxcp
->rss_hash
= GET_RX_COMPL_V0_BITS(rsshash
, compl);
2248 rxcp
->qnq
= GET_RX_COMPL_V0_BITS(qnq
, compl);
2249 rxcp
->vlan_tag
= GET_RX_COMPL_V0_BITS(vlan_tag
, compl);
2251 rxcp
->port
= GET_RX_COMPL_V0_BITS(port
, compl);
2252 rxcp
->ip_frag
= GET_RX_COMPL_V0_BITS(ip_frag
, compl);
2255 static struct be_rx_compl_info
*be_rx_compl_get(struct be_rx_obj
*rxo
)
2257 struct be_eth_rx_compl
*compl = queue_tail_node(&rxo
->cq
);
2258 struct be_rx_compl_info
*rxcp
= &rxo
->rxcp
;
2259 struct be_adapter
*adapter
= rxo
->adapter
;
2261 /* For checking the valid bit it is Ok to use either definition as the
2262 * valid bit is at the same position in both v0 and v1 Rx compl */
2263 if (compl->dw
[offsetof(struct amap_eth_rx_compl_v1
, valid
) / 32] == 0)
2267 be_dws_le_to_cpu(compl, sizeof(*compl));
2269 if (adapter
->be3_native
)
2270 be_parse_rx_compl_v1(compl, rxcp
);
2272 be_parse_rx_compl_v0(compl, rxcp
);
2278 /* In QNQ modes, if qnq bit is not set, then the packet was
2279 * tagged only with the transparent outer vlan-tag and must
2280 * not be treated as a vlan packet by host
2282 if (be_is_qnq_mode(adapter
) && !rxcp
->qnq
)
2285 if (!lancer_chip(adapter
))
2286 rxcp
->vlan_tag
= swab16(rxcp
->vlan_tag
);
2288 if (adapter
->pvid
== (rxcp
->vlan_tag
& VLAN_VID_MASK
) &&
2289 !test_bit(rxcp
->vlan_tag
, adapter
->vids
))
2293 /* As the compl has been parsed, reset it; we wont touch it again */
2294 compl->dw
[offsetof(struct amap_eth_rx_compl_v1
, valid
) / 32] = 0;
2296 queue_tail_inc(&rxo
->cq
);
2300 static inline struct page
*be_alloc_pages(u32 size
, gfp_t gfp
)
2302 u32 order
= get_order(size
);
2306 return alloc_pages(gfp
, order
);
2310 * Allocate a page, split it to fragments of size rx_frag_size and post as
2311 * receive buffers to BE
2313 static void be_post_rx_frags(struct be_rx_obj
*rxo
, gfp_t gfp
, u32 frags_needed
)
2315 struct be_adapter
*adapter
= rxo
->adapter
;
2316 struct be_rx_page_info
*page_info
= NULL
, *prev_page_info
= NULL
;
2317 struct be_queue_info
*rxq
= &rxo
->q
;
2318 struct page
*pagep
= NULL
;
2319 struct device
*dev
= &adapter
->pdev
->dev
;
2320 struct be_eth_rx_d
*rxd
;
2321 u64 page_dmaaddr
= 0, frag_dmaaddr
;
2322 u32 posted
, page_offset
= 0, notify
= 0;
2324 page_info
= &rxo
->page_info_tbl
[rxq
->head
];
2325 for (posted
= 0; posted
< frags_needed
&& !page_info
->page
; posted
++) {
2327 pagep
= be_alloc_pages(adapter
->big_page_size
, gfp
);
2328 if (unlikely(!pagep
)) {
2329 rx_stats(rxo
)->rx_post_fail
++;
2332 page_dmaaddr
= dma_map_page(dev
, pagep
, 0,
2333 adapter
->big_page_size
,
2335 if (dma_mapping_error(dev
, page_dmaaddr
)) {
2338 adapter
->drv_stats
.dma_map_errors
++;
2344 page_offset
+= rx_frag_size
;
2346 page_info
->page_offset
= page_offset
;
2347 page_info
->page
= pagep
;
2349 rxd
= queue_head_node(rxq
);
2350 frag_dmaaddr
= page_dmaaddr
+ page_info
->page_offset
;
2351 rxd
->fragpa_lo
= cpu_to_le32(frag_dmaaddr
& 0xFFFFFFFF);
2352 rxd
->fragpa_hi
= cpu_to_le32(upper_32_bits(frag_dmaaddr
));
2354 /* Any space left in the current big page for another frag? */
2355 if ((page_offset
+ rx_frag_size
+ rx_frag_size
) >
2356 adapter
->big_page_size
) {
2358 page_info
->last_frag
= true;
2359 dma_unmap_addr_set(page_info
, bus
, page_dmaaddr
);
2361 dma_unmap_addr_set(page_info
, bus
, frag_dmaaddr
);
2364 prev_page_info
= page_info
;
2365 queue_head_inc(rxq
);
2366 page_info
= &rxo
->page_info_tbl
[rxq
->head
];
2369 /* Mark the last frag of a page when we break out of the above loop
2370 * with no more slots available in the RXQ
2373 prev_page_info
->last_frag
= true;
2374 dma_unmap_addr_set(prev_page_info
, bus
, page_dmaaddr
);
2378 atomic_add(posted
, &rxq
->used
);
2379 if (rxo
->rx_post_starved
)
2380 rxo
->rx_post_starved
= false;
2382 notify
= min(MAX_NUM_POST_ERX_DB
, posted
);
2383 be_rxq_notify(adapter
, rxq
->id
, notify
);
2386 } else if (atomic_read(&rxq
->used
) == 0) {
2387 /* Let be_worker replenish when memory is available */
2388 rxo
->rx_post_starved
= true;
2392 static struct be_tx_compl_info
*be_tx_compl_get(struct be_tx_obj
*txo
)
2394 struct be_queue_info
*tx_cq
= &txo
->cq
;
2395 struct be_tx_compl_info
*txcp
= &txo
->txcp
;
2396 struct be_eth_tx_compl
*compl = queue_tail_node(tx_cq
);
2398 if (compl->dw
[offsetof(struct amap_eth_tx_compl
, valid
) / 32] == 0)
2401 /* Ensure load ordering of valid bit dword and other dwords below */
2403 be_dws_le_to_cpu(compl, sizeof(*compl));
2405 txcp
->status
= GET_TX_COMPL_BITS(status
, compl);
2406 txcp
->end_index
= GET_TX_COMPL_BITS(wrb_index
, compl);
2408 compl->dw
[offsetof(struct amap_eth_tx_compl
, valid
) / 32] = 0;
2409 queue_tail_inc(tx_cq
);
2413 static u16
be_tx_compl_process(struct be_adapter
*adapter
,
2414 struct be_tx_obj
*txo
, u16 last_index
)
2416 struct sk_buff
**sent_skbs
= txo
->sent_skb_list
;
2417 struct be_queue_info
*txq
= &txo
->q
;
2418 struct sk_buff
*skb
= NULL
;
2419 bool unmap_skb_hdr
= false;
2420 struct be_eth_wrb
*wrb
;
2425 if (sent_skbs
[txq
->tail
]) {
2426 /* Free skb from prev req */
2428 dev_consume_skb_any(skb
);
2429 skb
= sent_skbs
[txq
->tail
];
2430 sent_skbs
[txq
->tail
] = NULL
;
2431 queue_tail_inc(txq
); /* skip hdr wrb */
2433 unmap_skb_hdr
= true;
2435 wrb
= queue_tail_node(txq
);
2436 frag_index
= txq
->tail
;
2437 unmap_tx_frag(&adapter
->pdev
->dev
, wrb
,
2438 (unmap_skb_hdr
&& skb_headlen(skb
)));
2439 unmap_skb_hdr
= false;
2440 queue_tail_inc(txq
);
2442 } while (frag_index
!= last_index
);
2443 dev_consume_skb_any(skb
);
2448 /* Return the number of events in the event queue */
2449 static inline int events_get(struct be_eq_obj
*eqo
)
2451 struct be_eq_entry
*eqe
;
2455 eqe
= queue_tail_node(&eqo
->q
);
2462 queue_tail_inc(&eqo
->q
);
2468 /* Leaves the EQ is disarmed state */
2469 static void be_eq_clean(struct be_eq_obj
*eqo
)
2471 int num
= events_get(eqo
);
2473 be_eq_notify(eqo
->adapter
, eqo
->q
.id
, false, true, num
, 0);
2476 /* Free posted rx buffers that were not used */
2477 static void be_rxq_clean(struct be_rx_obj
*rxo
)
2479 struct be_queue_info
*rxq
= &rxo
->q
;
2480 struct be_rx_page_info
*page_info
;
2482 while (atomic_read(&rxq
->used
) > 0) {
2483 page_info
= get_rx_page_info(rxo
);
2484 put_page(page_info
->page
);
2485 memset(page_info
, 0, sizeof(*page_info
));
2487 BUG_ON(atomic_read(&rxq
->used
));
2492 static void be_rx_cq_clean(struct be_rx_obj
*rxo
)
2494 struct be_queue_info
*rx_cq
= &rxo
->cq
;
2495 struct be_rx_compl_info
*rxcp
;
2496 struct be_adapter
*adapter
= rxo
->adapter
;
2499 /* Consume pending rx completions.
2500 * Wait for the flush completion (identified by zero num_rcvd)
2501 * to arrive. Notify CQ even when there are no more CQ entries
2502 * for HW to flush partially coalesced CQ entries.
2503 * In Lancer, there is no need to wait for flush compl.
2506 rxcp
= be_rx_compl_get(rxo
);
2508 if (lancer_chip(adapter
))
2511 if (flush_wait
++ > 50 ||
2512 be_check_error(adapter
,
2514 dev_warn(&adapter
->pdev
->dev
,
2515 "did not receive flush compl\n");
2518 be_cq_notify(adapter
, rx_cq
->id
, true, 0);
2521 be_rx_compl_discard(rxo
, rxcp
);
2522 be_cq_notify(adapter
, rx_cq
->id
, false, 1);
2523 if (rxcp
->num_rcvd
== 0)
2528 /* After cleanup, leave the CQ in unarmed state */
2529 be_cq_notify(adapter
, rx_cq
->id
, false, 0);
2532 static void be_tx_compl_clean(struct be_adapter
*adapter
)
2534 struct device
*dev
= &adapter
->pdev
->dev
;
2535 u16 cmpl
= 0, timeo
= 0, num_wrbs
= 0;
2536 struct be_tx_compl_info
*txcp
;
2537 struct be_queue_info
*txq
;
2538 u32 end_idx
, notified_idx
;
2539 struct be_tx_obj
*txo
;
2540 int i
, pending_txqs
;
2542 /* Stop polling for compls when HW has been silent for 10ms */
2544 pending_txqs
= adapter
->num_tx_qs
;
2546 for_all_tx_queues(adapter
, txo
, i
) {
2550 while ((txcp
= be_tx_compl_get(txo
))) {
2552 be_tx_compl_process(adapter
, txo
,
2557 be_cq_notify(adapter
, txo
->cq
.id
, false, cmpl
);
2558 atomic_sub(num_wrbs
, &txq
->used
);
2561 if (!be_is_tx_compl_pending(txo
))
2565 if (pending_txqs
== 0 || ++timeo
> 10 ||
2566 be_check_error(adapter
, BE_ERROR_HW
))
2572 /* Free enqueued TX that was never notified to HW */
2573 for_all_tx_queues(adapter
, txo
, i
) {
2576 if (atomic_read(&txq
->used
)) {
2577 dev_info(dev
, "txq%d: cleaning %d pending tx-wrbs\n",
2578 i
, atomic_read(&txq
->used
));
2579 notified_idx
= txq
->tail
;
2580 end_idx
= txq
->tail
;
2581 index_adv(&end_idx
, atomic_read(&txq
->used
) - 1,
2583 /* Use the tx-compl process logic to handle requests
2584 * that were not sent to the HW.
2586 num_wrbs
= be_tx_compl_process(adapter
, txo
, end_idx
);
2587 atomic_sub(num_wrbs
, &txq
->used
);
2588 BUG_ON(atomic_read(&txq
->used
));
2589 txo
->pend_wrb_cnt
= 0;
2590 /* Since hw was never notified of these requests,
2593 txq
->head
= notified_idx
;
2594 txq
->tail
= notified_idx
;
2599 static void be_evt_queues_destroy(struct be_adapter
*adapter
)
2601 struct be_eq_obj
*eqo
;
2604 for_all_evt_queues(adapter
, eqo
, i
) {
2605 if (eqo
->q
.created
) {
2607 be_cmd_q_destroy(adapter
, &eqo
->q
, QTYPE_EQ
);
2608 napi_hash_del(&eqo
->napi
);
2609 netif_napi_del(&eqo
->napi
);
2610 free_cpumask_var(eqo
->affinity_mask
);
2612 be_queue_free(adapter
, &eqo
->q
);
2616 static int be_evt_queues_create(struct be_adapter
*adapter
)
2618 struct be_queue_info
*eq
;
2619 struct be_eq_obj
*eqo
;
2620 struct be_aic_obj
*aic
;
2623 adapter
->num_evt_qs
= min_t(u16
, num_irqs(adapter
),
2624 adapter
->cfg_num_qs
);
2626 for_all_evt_queues(adapter
, eqo
, i
) {
2627 int numa_node
= dev_to_node(&adapter
->pdev
->dev
);
2629 aic
= &adapter
->aic_obj
[i
];
2630 eqo
->adapter
= adapter
;
2632 aic
->max_eqd
= BE_MAX_EQD
;
2636 rc
= be_queue_alloc(adapter
, eq
, EVNT_Q_LEN
,
2637 sizeof(struct be_eq_entry
));
2641 rc
= be_cmd_eq_create(adapter
, eqo
);
2645 if (!zalloc_cpumask_var(&eqo
->affinity_mask
, GFP_KERNEL
))
2647 cpumask_set_cpu(cpumask_local_spread(i
, numa_node
),
2648 eqo
->affinity_mask
);
2649 netif_napi_add(adapter
->netdev
, &eqo
->napi
, be_poll
,
2655 static void be_mcc_queues_destroy(struct be_adapter
*adapter
)
2657 struct be_queue_info
*q
;
2659 q
= &adapter
->mcc_obj
.q
;
2661 be_cmd_q_destroy(adapter
, q
, QTYPE_MCCQ
);
2662 be_queue_free(adapter
, q
);
2664 q
= &adapter
->mcc_obj
.cq
;
2666 be_cmd_q_destroy(adapter
, q
, QTYPE_CQ
);
2667 be_queue_free(adapter
, q
);
2670 /* Must be called only after TX qs are created as MCC shares TX EQ */
2671 static int be_mcc_queues_create(struct be_adapter
*adapter
)
2673 struct be_queue_info
*q
, *cq
;
2675 cq
= &adapter
->mcc_obj
.cq
;
2676 if (be_queue_alloc(adapter
, cq
, MCC_CQ_LEN
,
2677 sizeof(struct be_mcc_compl
)))
2680 /* Use the default EQ for MCC completions */
2681 if (be_cmd_cq_create(adapter
, cq
, &mcc_eqo(adapter
)->q
, true, 0))
2684 q
= &adapter
->mcc_obj
.q
;
2685 if (be_queue_alloc(adapter
, q
, MCC_Q_LEN
, sizeof(struct be_mcc_wrb
)))
2686 goto mcc_cq_destroy
;
2688 if (be_cmd_mccq_create(adapter
, q
, cq
))
2694 be_queue_free(adapter
, q
);
2696 be_cmd_q_destroy(adapter
, cq
, QTYPE_CQ
);
2698 be_queue_free(adapter
, cq
);
2703 static void be_tx_queues_destroy(struct be_adapter
*adapter
)
2705 struct be_queue_info
*q
;
2706 struct be_tx_obj
*txo
;
2709 for_all_tx_queues(adapter
, txo
, i
) {
2712 be_cmd_q_destroy(adapter
, q
, QTYPE_TXQ
);
2713 be_queue_free(adapter
, q
);
2717 be_cmd_q_destroy(adapter
, q
, QTYPE_CQ
);
2718 be_queue_free(adapter
, q
);
2722 static int be_tx_qs_create(struct be_adapter
*adapter
)
2724 struct be_queue_info
*cq
;
2725 struct be_tx_obj
*txo
;
2726 struct be_eq_obj
*eqo
;
2729 adapter
->num_tx_qs
= min(adapter
->num_evt_qs
, be_max_txqs(adapter
));
2731 for_all_tx_queues(adapter
, txo
, i
) {
2733 status
= be_queue_alloc(adapter
, cq
, TX_CQ_LEN
,
2734 sizeof(struct be_eth_tx_compl
));
2738 u64_stats_init(&txo
->stats
.sync
);
2739 u64_stats_init(&txo
->stats
.sync_compl
);
2741 /* If num_evt_qs is less than num_tx_qs, then more than
2742 * one txq share an eq
2744 eqo
= &adapter
->eq_obj
[i
% adapter
->num_evt_qs
];
2745 status
= be_cmd_cq_create(adapter
, cq
, &eqo
->q
, false, 3);
2749 status
= be_queue_alloc(adapter
, &txo
->q
, TX_Q_LEN
,
2750 sizeof(struct be_eth_wrb
));
2754 status
= be_cmd_txq_create(adapter
, txo
);
2758 netif_set_xps_queue(adapter
->netdev
, eqo
->affinity_mask
,
2762 dev_info(&adapter
->pdev
->dev
, "created %d TX queue(s)\n",
2763 adapter
->num_tx_qs
);
2767 static void be_rx_cqs_destroy(struct be_adapter
*adapter
)
2769 struct be_queue_info
*q
;
2770 struct be_rx_obj
*rxo
;
2773 for_all_rx_queues(adapter
, rxo
, i
) {
2776 be_cmd_q_destroy(adapter
, q
, QTYPE_CQ
);
2777 be_queue_free(adapter
, q
);
2781 static int be_rx_cqs_create(struct be_adapter
*adapter
)
2783 struct be_queue_info
*eq
, *cq
;
2784 struct be_rx_obj
*rxo
;
2787 /* We can create as many RSS rings as there are EQs. */
2788 adapter
->num_rss_qs
= adapter
->num_evt_qs
;
2790 /* We'll use RSS only if atleast 2 RSS rings are supported. */
2791 if (adapter
->num_rss_qs
<= 1)
2792 adapter
->num_rss_qs
= 0;
2794 adapter
->num_rx_qs
= adapter
->num_rss_qs
+ adapter
->need_def_rxq
;
2796 /* When the interface is not capable of RSS rings (and there is no
2797 * need to create a default RXQ) we'll still need one RXQ
2799 if (adapter
->num_rx_qs
== 0)
2800 adapter
->num_rx_qs
= 1;
2802 adapter
->big_page_size
= (1 << get_order(rx_frag_size
)) * PAGE_SIZE
;
2803 for_all_rx_queues(adapter
, rxo
, i
) {
2804 rxo
->adapter
= adapter
;
2806 rc
= be_queue_alloc(adapter
, cq
, RX_CQ_LEN
,
2807 sizeof(struct be_eth_rx_compl
));
2811 u64_stats_init(&rxo
->stats
.sync
);
2812 eq
= &adapter
->eq_obj
[i
% adapter
->num_evt_qs
].q
;
2813 rc
= be_cmd_cq_create(adapter
, cq
, eq
, false, 3);
2818 dev_info(&adapter
->pdev
->dev
,
2819 "created %d RX queue(s)\n", adapter
->num_rx_qs
);
2823 static irqreturn_t
be_intx(int irq
, void *dev
)
2825 struct be_eq_obj
*eqo
= dev
;
2826 struct be_adapter
*adapter
= eqo
->adapter
;
2829 /* IRQ is not expected when NAPI is scheduled as the EQ
2830 * will not be armed.
2831 * But, this can happen on Lancer INTx where it takes
2832 * a while to de-assert INTx or in BE2 where occasionaly
2833 * an interrupt may be raised even when EQ is unarmed.
2834 * If NAPI is already scheduled, then counting & notifying
2835 * events will orphan them.
2837 if (napi_schedule_prep(&eqo
->napi
)) {
2838 num_evts
= events_get(eqo
);
2839 __napi_schedule(&eqo
->napi
);
2841 eqo
->spurious_intr
= 0;
2843 be_eq_notify(adapter
, eqo
->q
.id
, false, true, num_evts
, 0);
2845 /* Return IRQ_HANDLED only for the the first spurious intr
2846 * after a valid intr to stop the kernel from branding
2847 * this irq as a bad one!
2849 if (num_evts
|| eqo
->spurious_intr
++ == 0)
2855 static irqreturn_t
be_msix(int irq
, void *dev
)
2857 struct be_eq_obj
*eqo
= dev
;
2859 be_eq_notify(eqo
->adapter
, eqo
->q
.id
, false, true, 0, 0);
2860 napi_schedule(&eqo
->napi
);
2864 static inline bool do_gro(struct be_rx_compl_info
*rxcp
)
2866 return (rxcp
->tcpf
&& !rxcp
->err
&& rxcp
->l4_csum
) ? true : false;
2869 static int be_process_rx(struct be_rx_obj
*rxo
, struct napi_struct
*napi
,
2870 int budget
, int polling
)
2872 struct be_adapter
*adapter
= rxo
->adapter
;
2873 struct be_queue_info
*rx_cq
= &rxo
->cq
;
2874 struct be_rx_compl_info
*rxcp
;
2876 u32 frags_consumed
= 0;
2878 for (work_done
= 0; work_done
< budget
; work_done
++) {
2879 rxcp
= be_rx_compl_get(rxo
);
2883 /* Is it a flush compl that has no data */
2884 if (unlikely(rxcp
->num_rcvd
== 0))
2887 /* Discard compl with partial DMA Lancer B0 */
2888 if (unlikely(!rxcp
->pkt_size
)) {
2889 be_rx_compl_discard(rxo
, rxcp
);
2893 /* On BE drop pkts that arrive due to imperfect filtering in
2894 * promiscuous mode on some skews
2896 if (unlikely(rxcp
->port
!= adapter
->port_num
&&
2897 !lancer_chip(adapter
))) {
2898 be_rx_compl_discard(rxo
, rxcp
);
2902 /* Don't do gro when we're busy_polling */
2903 if (do_gro(rxcp
) && polling
!= BUSY_POLLING
)
2904 be_rx_compl_process_gro(rxo
, napi
, rxcp
);
2906 be_rx_compl_process(rxo
, napi
, rxcp
);
2909 frags_consumed
+= rxcp
->num_rcvd
;
2910 be_rx_stats_update(rxo
, rxcp
);
2914 be_cq_notify(adapter
, rx_cq
->id
, true, work_done
);
2916 /* When an rx-obj gets into post_starved state, just
2917 * let be_worker do the posting.
2919 if (atomic_read(&rxo
->q
.used
) < RX_FRAGS_REFILL_WM
&&
2920 !rxo
->rx_post_starved
)
2921 be_post_rx_frags(rxo
, GFP_ATOMIC
,
2922 max_t(u32
, MAX_RX_POST
,
2929 static inline void be_update_tx_err(struct be_tx_obj
*txo
, u8 status
)
2932 case BE_TX_COMP_HDR_PARSE_ERR
:
2933 tx_stats(txo
)->tx_hdr_parse_err
++;
2935 case BE_TX_COMP_NDMA_ERR
:
2936 tx_stats(txo
)->tx_dma_err
++;
2938 case BE_TX_COMP_ACL_ERR
:
2939 tx_stats(txo
)->tx_spoof_check_err
++;
2944 static inline void lancer_update_tx_err(struct be_tx_obj
*txo
, u8 status
)
2947 case LANCER_TX_COMP_LSO_ERR
:
2948 tx_stats(txo
)->tx_tso_err
++;
2950 case LANCER_TX_COMP_HSW_DROP_MAC_ERR
:
2951 case LANCER_TX_COMP_HSW_DROP_VLAN_ERR
:
2952 tx_stats(txo
)->tx_spoof_check_err
++;
2954 case LANCER_TX_COMP_QINQ_ERR
:
2955 tx_stats(txo
)->tx_qinq_err
++;
2957 case LANCER_TX_COMP_PARITY_ERR
:
2958 tx_stats(txo
)->tx_internal_parity_err
++;
2960 case LANCER_TX_COMP_DMA_ERR
:
2961 tx_stats(txo
)->tx_dma_err
++;
2966 static void be_process_tx(struct be_adapter
*adapter
, struct be_tx_obj
*txo
,
2969 int num_wrbs
= 0, work_done
= 0;
2970 struct be_tx_compl_info
*txcp
;
2972 while ((txcp
= be_tx_compl_get(txo
))) {
2973 num_wrbs
+= be_tx_compl_process(adapter
, txo
, txcp
->end_index
);
2977 if (lancer_chip(adapter
))
2978 lancer_update_tx_err(txo
, txcp
->status
);
2980 be_update_tx_err(txo
, txcp
->status
);
2985 be_cq_notify(adapter
, txo
->cq
.id
, true, work_done
);
2986 atomic_sub(num_wrbs
, &txo
->q
.used
);
2988 /* As Tx wrbs have been freed up, wake up netdev queue
2989 * if it was stopped due to lack of tx wrbs. */
2990 if (__netif_subqueue_stopped(adapter
->netdev
, idx
) &&
2991 be_can_txq_wake(txo
)) {
2992 netif_wake_subqueue(adapter
->netdev
, idx
);
2995 u64_stats_update_begin(&tx_stats(txo
)->sync_compl
);
2996 tx_stats(txo
)->tx_compl
+= work_done
;
2997 u64_stats_update_end(&tx_stats(txo
)->sync_compl
);
3001 #ifdef CONFIG_NET_RX_BUSY_POLL
3002 static inline bool be_lock_napi(struct be_eq_obj
*eqo
)
3006 spin_lock(&eqo
->lock
); /* BH is already disabled */
3007 if (eqo
->state
& BE_EQ_LOCKED
) {
3008 WARN_ON(eqo
->state
& BE_EQ_NAPI
);
3009 eqo
->state
|= BE_EQ_NAPI_YIELD
;
3012 eqo
->state
= BE_EQ_NAPI
;
3014 spin_unlock(&eqo
->lock
);
3018 static inline void be_unlock_napi(struct be_eq_obj
*eqo
)
3020 spin_lock(&eqo
->lock
); /* BH is already disabled */
3022 WARN_ON(eqo
->state
& (BE_EQ_POLL
| BE_EQ_NAPI_YIELD
));
3023 eqo
->state
= BE_EQ_IDLE
;
3025 spin_unlock(&eqo
->lock
);
3028 static inline bool be_lock_busy_poll(struct be_eq_obj
*eqo
)
3032 spin_lock_bh(&eqo
->lock
);
3033 if (eqo
->state
& BE_EQ_LOCKED
) {
3034 eqo
->state
|= BE_EQ_POLL_YIELD
;
3037 eqo
->state
|= BE_EQ_POLL
;
3039 spin_unlock_bh(&eqo
->lock
);
3043 static inline void be_unlock_busy_poll(struct be_eq_obj
*eqo
)
3045 spin_lock_bh(&eqo
->lock
);
3047 WARN_ON(eqo
->state
& (BE_EQ_NAPI
));
3048 eqo
->state
= BE_EQ_IDLE
;
3050 spin_unlock_bh(&eqo
->lock
);
3053 static inline void be_enable_busy_poll(struct be_eq_obj
*eqo
)
3055 spin_lock_init(&eqo
->lock
);
3056 eqo
->state
= BE_EQ_IDLE
;
3059 static inline void be_disable_busy_poll(struct be_eq_obj
*eqo
)
3063 /* It's enough to just acquire napi lock on the eqo to stop
3064 * be_busy_poll() from processing any queueus.
3066 while (!be_lock_napi(eqo
))
3072 #else /* CONFIG_NET_RX_BUSY_POLL */
3074 static inline bool be_lock_napi(struct be_eq_obj
*eqo
)
3079 static inline void be_unlock_napi(struct be_eq_obj
*eqo
)
3083 static inline bool be_lock_busy_poll(struct be_eq_obj
*eqo
)
3088 static inline void be_unlock_busy_poll(struct be_eq_obj
*eqo
)
3092 static inline void be_enable_busy_poll(struct be_eq_obj
*eqo
)
3096 static inline void be_disable_busy_poll(struct be_eq_obj
*eqo
)
3099 #endif /* CONFIG_NET_RX_BUSY_POLL */
3101 int be_poll(struct napi_struct
*napi
, int budget
)
3103 struct be_eq_obj
*eqo
= container_of(napi
, struct be_eq_obj
, napi
);
3104 struct be_adapter
*adapter
= eqo
->adapter
;
3105 int max_work
= 0, work
, i
, num_evts
;
3106 struct be_rx_obj
*rxo
;
3107 struct be_tx_obj
*txo
;
3110 num_evts
= events_get(eqo
);
3112 for_all_tx_queues_on_eq(adapter
, eqo
, txo
, i
)
3113 be_process_tx(adapter
, txo
, i
);
3115 if (be_lock_napi(eqo
)) {
3116 /* This loop will iterate twice for EQ0 in which
3117 * completions of the last RXQ (default one) are also processed
3118 * For other EQs the loop iterates only once
3120 for_all_rx_queues_on_eq(adapter
, eqo
, rxo
, i
) {
3121 work
= be_process_rx(rxo
, napi
, budget
, NAPI_POLLING
);
3122 max_work
= max(work
, max_work
);
3124 be_unlock_napi(eqo
);
3129 if (is_mcc_eqo(eqo
))
3130 be_process_mcc(adapter
);
3132 if (max_work
< budget
) {
3133 napi_complete(napi
);
3135 /* Skyhawk EQ_DB has a provision to set the rearm to interrupt
3136 * delay via a delay multiplier encoding value
3138 if (skyhawk_chip(adapter
))
3139 mult_enc
= be_get_eq_delay_mult_enc(eqo
);
3141 be_eq_notify(adapter
, eqo
->q
.id
, true, false, num_evts
,
3144 /* As we'll continue in polling mode, count and clear events */
3145 be_eq_notify(adapter
, eqo
->q
.id
, false, false, num_evts
, 0);
3150 #ifdef CONFIG_NET_RX_BUSY_POLL
3151 static int be_busy_poll(struct napi_struct
*napi
)
3153 struct be_eq_obj
*eqo
= container_of(napi
, struct be_eq_obj
, napi
);
3154 struct be_adapter
*adapter
= eqo
->adapter
;
3155 struct be_rx_obj
*rxo
;
3158 if (!be_lock_busy_poll(eqo
))
3159 return LL_FLUSH_BUSY
;
3161 for_all_rx_queues_on_eq(adapter
, eqo
, rxo
, i
) {
3162 work
= be_process_rx(rxo
, napi
, 4, BUSY_POLLING
);
3167 be_unlock_busy_poll(eqo
);
3172 void be_detect_error(struct be_adapter
*adapter
)
3174 u32 ue_lo
= 0, ue_hi
= 0, ue_lo_mask
= 0, ue_hi_mask
= 0;
3175 u32 sliport_status
= 0, sliport_err1
= 0, sliport_err2
= 0;
3177 struct device
*dev
= &adapter
->pdev
->dev
;
3179 if (be_check_error(adapter
, BE_ERROR_HW
))
3182 if (lancer_chip(adapter
)) {
3183 sliport_status
= ioread32(adapter
->db
+ SLIPORT_STATUS_OFFSET
);
3184 if (sliport_status
& SLIPORT_STATUS_ERR_MASK
) {
3185 be_set_error(adapter
, BE_ERROR_UE
);
3186 sliport_err1
= ioread32(adapter
->db
+
3187 SLIPORT_ERROR1_OFFSET
);
3188 sliport_err2
= ioread32(adapter
->db
+
3189 SLIPORT_ERROR2_OFFSET
);
3190 /* Do not log error messages if its a FW reset */
3191 if (sliport_err1
== SLIPORT_ERROR_FW_RESET1
&&
3192 sliport_err2
== SLIPORT_ERROR_FW_RESET2
) {
3193 dev_info(dev
, "Firmware update in progress\n");
3195 dev_err(dev
, "Error detected in the card\n");
3196 dev_err(dev
, "ERR: sliport status 0x%x\n",
3198 dev_err(dev
, "ERR: sliport error1 0x%x\n",
3200 dev_err(dev
, "ERR: sliport error2 0x%x\n",
3205 ue_lo
= ioread32(adapter
->pcicfg
+ PCICFG_UE_STATUS_LOW
);
3206 ue_hi
= ioread32(adapter
->pcicfg
+ PCICFG_UE_STATUS_HIGH
);
3207 ue_lo_mask
= ioread32(adapter
->pcicfg
+
3208 PCICFG_UE_STATUS_LOW_MASK
);
3209 ue_hi_mask
= ioread32(adapter
->pcicfg
+
3210 PCICFG_UE_STATUS_HI_MASK
);
3212 ue_lo
= (ue_lo
& ~ue_lo_mask
);
3213 ue_hi
= (ue_hi
& ~ue_hi_mask
);
3215 /* On certain platforms BE hardware can indicate spurious UEs.
3216 * Allow HW to stop working completely in case of a real UE.
3217 * Hence not setting the hw_error for UE detection.
3220 if (ue_lo
|| ue_hi
) {
3222 "Unrecoverable Error detected in the adapter");
3223 dev_err(dev
, "Please reboot server to recover");
3224 if (skyhawk_chip(adapter
))
3225 be_set_error(adapter
, BE_ERROR_UE
);
3227 for (i
= 0; ue_lo
; ue_lo
>>= 1, i
++) {
3229 dev_err(dev
, "UE: %s bit set\n",
3230 ue_status_low_desc
[i
]);
3232 for (i
= 0; ue_hi
; ue_hi
>>= 1, i
++) {
3234 dev_err(dev
, "UE: %s bit set\n",
3235 ue_status_hi_desc
[i
]);
3241 static void be_msix_disable(struct be_adapter
*adapter
)
3243 if (msix_enabled(adapter
)) {
3244 pci_disable_msix(adapter
->pdev
);
3245 adapter
->num_msix_vec
= 0;
3246 adapter
->num_msix_roce_vec
= 0;
3250 static int be_msix_enable(struct be_adapter
*adapter
)
3253 struct device
*dev
= &adapter
->pdev
->dev
;
3255 /* If RoCE is supported, program the max number of NIC vectors that
3256 * may be configured via set-channels, along with vectors needed for
3257 * RoCe. Else, just program the number we'll use initially.
3259 if (be_roce_supported(adapter
))
3260 num_vec
= min_t(int, 2 * be_max_eqs(adapter
),
3261 2 * num_online_cpus());
3263 num_vec
= adapter
->cfg_num_qs
;
3265 for (i
= 0; i
< num_vec
; i
++)
3266 adapter
->msix_entries
[i
].entry
= i
;
3268 num_vec
= pci_enable_msix_range(adapter
->pdev
, adapter
->msix_entries
,
3269 MIN_MSIX_VECTORS
, num_vec
);
3273 if (be_roce_supported(adapter
) && num_vec
> MIN_MSIX_VECTORS
) {
3274 adapter
->num_msix_roce_vec
= num_vec
/ 2;
3275 dev_info(dev
, "enabled %d MSI-x vector(s) for RoCE\n",
3276 adapter
->num_msix_roce_vec
);
3279 adapter
->num_msix_vec
= num_vec
- adapter
->num_msix_roce_vec
;
3281 dev_info(dev
, "enabled %d MSI-x vector(s) for NIC\n",
3282 adapter
->num_msix_vec
);
3286 dev_warn(dev
, "MSIx enable failed\n");
3288 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
3289 if (be_virtfn(adapter
))
3294 static inline int be_msix_vec_get(struct be_adapter
*adapter
,
3295 struct be_eq_obj
*eqo
)
3297 return adapter
->msix_entries
[eqo
->msix_idx
].vector
;
3300 static int be_msix_register(struct be_adapter
*adapter
)
3302 struct net_device
*netdev
= adapter
->netdev
;
3303 struct be_eq_obj
*eqo
;
3306 for_all_evt_queues(adapter
, eqo
, i
) {
3307 sprintf(eqo
->desc
, "%s-q%d", netdev
->name
, i
);
3308 vec
= be_msix_vec_get(adapter
, eqo
);
3309 status
= request_irq(vec
, be_msix
, 0, eqo
->desc
, eqo
);
3313 irq_set_affinity_hint(vec
, eqo
->affinity_mask
);
3318 for (i
--; i
>= 0; i
--) {
3319 eqo
= &adapter
->eq_obj
[i
];
3320 free_irq(be_msix_vec_get(adapter
, eqo
), eqo
);
3322 dev_warn(&adapter
->pdev
->dev
, "MSIX Request IRQ failed - err %d\n",
3324 be_msix_disable(adapter
);
3328 static int be_irq_register(struct be_adapter
*adapter
)
3330 struct net_device
*netdev
= adapter
->netdev
;
3333 if (msix_enabled(adapter
)) {
3334 status
= be_msix_register(adapter
);
3337 /* INTx is not supported for VF */
3338 if (be_virtfn(adapter
))
3342 /* INTx: only the first EQ is used */
3343 netdev
->irq
= adapter
->pdev
->irq
;
3344 status
= request_irq(netdev
->irq
, be_intx
, IRQF_SHARED
, netdev
->name
,
3345 &adapter
->eq_obj
[0]);
3347 dev_err(&adapter
->pdev
->dev
,
3348 "INTx request IRQ failed - err %d\n", status
);
3352 adapter
->isr_registered
= true;
3356 static void be_irq_unregister(struct be_adapter
*adapter
)
3358 struct net_device
*netdev
= adapter
->netdev
;
3359 struct be_eq_obj
*eqo
;
3362 if (!adapter
->isr_registered
)
3366 if (!msix_enabled(adapter
)) {
3367 free_irq(netdev
->irq
, &adapter
->eq_obj
[0]);
3372 for_all_evt_queues(adapter
, eqo
, i
) {
3373 vec
= be_msix_vec_get(adapter
, eqo
);
3374 irq_set_affinity_hint(vec
, NULL
);
3379 adapter
->isr_registered
= false;
3382 static void be_rx_qs_destroy(struct be_adapter
*adapter
)
3384 struct rss_info
*rss
= &adapter
->rss_info
;
3385 struct be_queue_info
*q
;
3386 struct be_rx_obj
*rxo
;
3389 for_all_rx_queues(adapter
, rxo
, i
) {
3392 /* If RXQs are destroyed while in an "out of buffer"
3393 * state, there is a possibility of an HW stall on
3394 * Lancer. So, post 64 buffers to each queue to relieve
3395 * the "out of buffer" condition.
3396 * Make sure there's space in the RXQ before posting.
3398 if (lancer_chip(adapter
)) {
3399 be_rx_cq_clean(rxo
);
3400 if (atomic_read(&q
->used
) == 0)
3401 be_post_rx_frags(rxo
, GFP_KERNEL
,
3405 be_cmd_rxq_destroy(adapter
, q
);
3406 be_rx_cq_clean(rxo
);
3409 be_queue_free(adapter
, q
);
3412 if (rss
->rss_flags
) {
3413 rss
->rss_flags
= RSS_ENABLE_NONE
;
3414 be_cmd_rss_config(adapter
, rss
->rsstable
, rss
->rss_flags
,
3415 128, rss
->rss_hkey
);
3419 static void be_disable_if_filters(struct be_adapter
*adapter
)
3421 be_cmd_pmac_del(adapter
, adapter
->if_handle
,
3422 adapter
->pmac_id
[0], 0);
3424 be_clear_uc_list(adapter
);
3426 /* The IFACE flags are enabled in the open path and cleared
3427 * in the close path. When a VF gets detached from the host and
3428 * assigned to a VM the following happens:
3429 * - VF's IFACE flags get cleared in the detach path
3430 * - IFACE create is issued by the VF in the attach path
3431 * Due to a bug in the BE3/Skyhawk-R FW
3432 * (Lancer FW doesn't have the bug), the IFACE capability flags
3433 * specified along with the IFACE create cmd issued by a VF are not
3434 * honoured by FW. As a consequence, if a *new* driver
3435 * (that enables/disables IFACE flags in open/close)
3436 * is loaded in the host and an *old* driver is * used by a VM/VF,
3437 * the IFACE gets created *without* the needed flags.
3438 * To avoid this, disable RX-filter flags only for Lancer.
3440 if (lancer_chip(adapter
)) {
3441 be_cmd_rx_filter(adapter
, BE_IF_ALL_FILT_FLAGS
, OFF
);
3442 adapter
->if_flags
&= ~BE_IF_ALL_FILT_FLAGS
;
3446 static int be_close(struct net_device
*netdev
)
3448 struct be_adapter
*adapter
= netdev_priv(netdev
);
3449 struct be_eq_obj
*eqo
;
3452 /* This protection is needed as be_close() may be called even when the
3453 * adapter is in cleared state (after eeh perm failure)
3455 if (!(adapter
->flags
& BE_FLAGS_SETUP_DONE
))
3458 be_disable_if_filters(adapter
);
3460 if (adapter
->flags
& BE_FLAGS_NAPI_ENABLED
) {
3461 for_all_evt_queues(adapter
, eqo
, i
) {
3462 napi_disable(&eqo
->napi
);
3463 be_disable_busy_poll(eqo
);
3465 adapter
->flags
&= ~BE_FLAGS_NAPI_ENABLED
;
3468 be_async_mcc_disable(adapter
);
3470 /* Wait for all pending tx completions to arrive so that
3471 * all tx skbs are freed.
3473 netif_tx_disable(netdev
);
3474 be_tx_compl_clean(adapter
);
3476 be_rx_qs_destroy(adapter
);
3478 for_all_evt_queues(adapter
, eqo
, i
) {
3479 if (msix_enabled(adapter
))
3480 synchronize_irq(be_msix_vec_get(adapter
, eqo
));
3482 synchronize_irq(netdev
->irq
);
3486 be_irq_unregister(adapter
);
3491 static int be_rx_qs_create(struct be_adapter
*adapter
)
3493 struct rss_info
*rss
= &adapter
->rss_info
;
3494 u8 rss_key
[RSS_HASH_KEY_LEN
];
3495 struct be_rx_obj
*rxo
;
3498 for_all_rx_queues(adapter
, rxo
, i
) {
3499 rc
= be_queue_alloc(adapter
, &rxo
->q
, RX_Q_LEN
,
3500 sizeof(struct be_eth_rx_d
));
3505 if (adapter
->need_def_rxq
|| !adapter
->num_rss_qs
) {
3506 rxo
= default_rxo(adapter
);
3507 rc
= be_cmd_rxq_create(adapter
, &rxo
->q
, rxo
->cq
.id
,
3508 rx_frag_size
, adapter
->if_handle
,
3509 false, &rxo
->rss_id
);
3514 for_all_rss_queues(adapter
, rxo
, i
) {
3515 rc
= be_cmd_rxq_create(adapter
, &rxo
->q
, rxo
->cq
.id
,
3516 rx_frag_size
, adapter
->if_handle
,
3517 true, &rxo
->rss_id
);
3522 if (be_multi_rxq(adapter
)) {
3523 for (j
= 0; j
< RSS_INDIR_TABLE_LEN
; j
+= adapter
->num_rss_qs
) {
3524 for_all_rss_queues(adapter
, rxo
, i
) {
3525 if ((j
+ i
) >= RSS_INDIR_TABLE_LEN
)
3527 rss
->rsstable
[j
+ i
] = rxo
->rss_id
;
3528 rss
->rss_queue
[j
+ i
] = i
;
3531 rss
->rss_flags
= RSS_ENABLE_TCP_IPV4
| RSS_ENABLE_IPV4
|
3532 RSS_ENABLE_TCP_IPV6
| RSS_ENABLE_IPV6
;
3534 if (!BEx_chip(adapter
))
3535 rss
->rss_flags
|= RSS_ENABLE_UDP_IPV4
|
3536 RSS_ENABLE_UDP_IPV6
;
3538 netdev_rss_key_fill(rss_key
, RSS_HASH_KEY_LEN
);
3539 rc
= be_cmd_rss_config(adapter
, rss
->rsstable
, rss
->rss_flags
,
3540 RSS_INDIR_TABLE_LEN
, rss_key
);
3542 rss
->rss_flags
= RSS_ENABLE_NONE
;
3546 memcpy(rss
->rss_hkey
, rss_key
, RSS_HASH_KEY_LEN
);
3548 /* Disable RSS, if only default RX Q is created */
3549 rss
->rss_flags
= RSS_ENABLE_NONE
;
3553 /* Post 1 less than RXQ-len to avoid head being equal to tail,
3554 * which is a queue empty condition
3556 for_all_rx_queues(adapter
, rxo
, i
)
3557 be_post_rx_frags(rxo
, GFP_KERNEL
, RX_Q_LEN
- 1);
3562 static int be_enable_if_filters(struct be_adapter
*adapter
)
3566 status
= be_cmd_rx_filter(adapter
, BE_IF_FILT_FLAGS_BASIC
, ON
);
3570 /* For BE3 VFs, the PF programs the initial MAC address */
3571 if (!(BEx_chip(adapter
) && be_virtfn(adapter
))) {
3572 status
= be_cmd_pmac_add(adapter
, adapter
->netdev
->dev_addr
,
3574 &adapter
->pmac_id
[0], 0);
3579 if (adapter
->vlans_added
)
3580 be_vid_config(adapter
);
3582 be_set_rx_mode(adapter
->netdev
);
3587 static int be_open(struct net_device
*netdev
)
3589 struct be_adapter
*adapter
= netdev_priv(netdev
);
3590 struct be_eq_obj
*eqo
;
3591 struct be_rx_obj
*rxo
;
3592 struct be_tx_obj
*txo
;
3596 status
= be_rx_qs_create(adapter
);
3600 status
= be_enable_if_filters(adapter
);
3604 status
= be_irq_register(adapter
);
3608 for_all_rx_queues(adapter
, rxo
, i
)
3609 be_cq_notify(adapter
, rxo
->cq
.id
, true, 0);
3611 for_all_tx_queues(adapter
, txo
, i
)
3612 be_cq_notify(adapter
, txo
->cq
.id
, true, 0);
3614 be_async_mcc_enable(adapter
);
3616 for_all_evt_queues(adapter
, eqo
, i
) {
3617 napi_enable(&eqo
->napi
);
3618 be_enable_busy_poll(eqo
);
3619 be_eq_notify(adapter
, eqo
->q
.id
, true, true, 0, 0);
3621 adapter
->flags
|= BE_FLAGS_NAPI_ENABLED
;
3623 status
= be_cmd_link_status_query(adapter
, NULL
, &link_status
, 0);
3625 be_link_status_update(adapter
, link_status
);
3627 netif_tx_start_all_queues(netdev
);
3628 #ifdef CONFIG_BE2NET_VXLAN
3629 if (skyhawk_chip(adapter
))
3630 vxlan_get_rx_port(netdev
);
3635 be_close(adapter
->netdev
);
3639 static int be_setup_wol(struct be_adapter
*adapter
, bool enable
)
3641 struct device
*dev
= &adapter
->pdev
->dev
;
3642 struct be_dma_mem cmd
;
3648 cmd
.size
= sizeof(struct be_cmd_req_acpi_wol_magic_config
);
3649 cmd
.va
= dma_zalloc_coherent(dev
, cmd
.size
, &cmd
.dma
, GFP_KERNEL
);
3654 status
= pci_write_config_dword(adapter
->pdev
,
3655 PCICFG_PM_CONTROL_OFFSET
,
3656 PCICFG_PM_CONTROL_MASK
);
3658 dev_err(dev
, "Could not enable Wake-on-lan\n");
3662 ether_addr_copy(mac
, adapter
->netdev
->dev_addr
);
3665 status
= be_cmd_enable_magic_wol(adapter
, mac
, &cmd
);
3666 pci_enable_wake(adapter
->pdev
, PCI_D3hot
, enable
);
3667 pci_enable_wake(adapter
->pdev
, PCI_D3cold
, enable
);
3669 dma_free_coherent(dev
, cmd
.size
, cmd
.va
, cmd
.dma
);
3673 static void be_vf_eth_addr_generate(struct be_adapter
*adapter
, u8
*mac
)
3677 addr
= jhash(adapter
->netdev
->dev_addr
, ETH_ALEN
, 0);
3679 mac
[5] = (u8
)(addr
& 0xFF);
3680 mac
[4] = (u8
)((addr
>> 8) & 0xFF);
3681 mac
[3] = (u8
)((addr
>> 16) & 0xFF);
3682 /* Use the OUI from the current MAC address */
3683 memcpy(mac
, adapter
->netdev
->dev_addr
, 3);
3687 * Generate a seed MAC address from the PF MAC Address using jhash.
3688 * MAC Address for VFs are assigned incrementally starting from the seed.
3689 * These addresses are programmed in the ASIC by the PF and the VF driver
3690 * queries for the MAC address during its probe.
3692 static int be_vf_eth_addr_config(struct be_adapter
*adapter
)
3697 struct be_vf_cfg
*vf_cfg
;
3699 be_vf_eth_addr_generate(adapter
, mac
);
3701 for_all_vfs(adapter
, vf_cfg
, vf
) {
3702 if (BEx_chip(adapter
))
3703 status
= be_cmd_pmac_add(adapter
, mac
,
3705 &vf_cfg
->pmac_id
, vf
+ 1);
3707 status
= be_cmd_set_mac(adapter
, mac
, vf_cfg
->if_handle
,
3711 dev_err(&adapter
->pdev
->dev
,
3712 "Mac address assignment failed for VF %d\n",
3715 memcpy(vf_cfg
->mac_addr
, mac
, ETH_ALEN
);
3722 static int be_vfs_mac_query(struct be_adapter
*adapter
)
3726 struct be_vf_cfg
*vf_cfg
;
3728 for_all_vfs(adapter
, vf_cfg
, vf
) {
3729 status
= be_cmd_get_active_mac(adapter
, vf_cfg
->pmac_id
,
3730 mac
, vf_cfg
->if_handle
,
3734 memcpy(vf_cfg
->mac_addr
, mac
, ETH_ALEN
);
3739 static void be_vf_clear(struct be_adapter
*adapter
)
3741 struct be_vf_cfg
*vf_cfg
;
3744 if (pci_vfs_assigned(adapter
->pdev
)) {
3745 dev_warn(&adapter
->pdev
->dev
,
3746 "VFs are assigned to VMs: not disabling VFs\n");
3750 pci_disable_sriov(adapter
->pdev
);
3752 for_all_vfs(adapter
, vf_cfg
, vf
) {
3753 if (BEx_chip(adapter
))
3754 be_cmd_pmac_del(adapter
, vf_cfg
->if_handle
,
3755 vf_cfg
->pmac_id
, vf
+ 1);
3757 be_cmd_set_mac(adapter
, NULL
, vf_cfg
->if_handle
,
3760 be_cmd_if_destroy(adapter
, vf_cfg
->if_handle
, vf
+ 1);
3763 kfree(adapter
->vf_cfg
);
3764 adapter
->num_vfs
= 0;
3765 adapter
->flags
&= ~BE_FLAGS_SRIOV_ENABLED
;
3768 static void be_clear_queues(struct be_adapter
*adapter
)
3770 be_mcc_queues_destroy(adapter
);
3771 be_rx_cqs_destroy(adapter
);
3772 be_tx_queues_destroy(adapter
);
3773 be_evt_queues_destroy(adapter
);
3776 static void be_cancel_worker(struct be_adapter
*adapter
)
3778 if (adapter
->flags
& BE_FLAGS_WORKER_SCHEDULED
) {
3779 cancel_delayed_work_sync(&adapter
->work
);
3780 adapter
->flags
&= ~BE_FLAGS_WORKER_SCHEDULED
;
3784 static void be_cancel_err_detection(struct be_adapter
*adapter
)
3786 if (adapter
->flags
& BE_FLAGS_ERR_DETECTION_SCHEDULED
) {
3787 cancel_delayed_work_sync(&adapter
->be_err_detection_work
);
3788 adapter
->flags
&= ~BE_FLAGS_ERR_DETECTION_SCHEDULED
;
3792 #ifdef CONFIG_BE2NET_VXLAN
3793 static void be_disable_vxlan_offloads(struct be_adapter
*adapter
)
3795 struct net_device
*netdev
= adapter
->netdev
;
3797 if (adapter
->flags
& BE_FLAGS_VXLAN_OFFLOADS
)
3798 be_cmd_manage_iface(adapter
, adapter
->if_handle
,
3799 OP_CONVERT_TUNNEL_TO_NORMAL
);
3801 if (adapter
->vxlan_port
)
3802 be_cmd_set_vxlan_port(adapter
, 0);
3804 adapter
->flags
&= ~BE_FLAGS_VXLAN_OFFLOADS
;
3805 adapter
->vxlan_port
= 0;
3807 netdev
->hw_enc_features
= 0;
3808 netdev
->hw_features
&= ~(NETIF_F_GSO_UDP_TUNNEL
);
3809 netdev
->features
&= ~(NETIF_F_GSO_UDP_TUNNEL
);
3813 static u16
be_calculate_vf_qs(struct be_adapter
*adapter
, u16 num_vfs
)
3815 struct be_resources res
= adapter
->pool_res
;
3818 /* Distribute the queue resources among the PF and it's VFs
3819 * Do not distribute queue resources in multi-channel configuration.
3821 if (num_vfs
&& !be_is_mc(adapter
)) {
3822 /* Divide the qpairs evenly among the VFs and the PF, capped
3823 * at VF-EQ-count. Any remainder qpairs belong to the PF.
3825 num_vf_qs
= min(SH_VF_MAX_NIC_EQS
,
3826 res
.max_rss_qs
/ (num_vfs
+ 1));
3828 /* Skyhawk-R chip supports only MAX_RSS_IFACES RSS capable
3829 * interfaces per port. Provide RSS on VFs, only if number
3830 * of VFs requested is less than MAX_RSS_IFACES limit.
3832 if (num_vfs
>= MAX_RSS_IFACES
)
3838 static int be_clear(struct be_adapter
*adapter
)
3840 struct pci_dev
*pdev
= adapter
->pdev
;
3843 be_cancel_worker(adapter
);
3845 if (sriov_enabled(adapter
))
3846 be_vf_clear(adapter
);
3848 /* Re-configure FW to distribute resources evenly across max-supported
3849 * number of VFs, only when VFs are not already enabled.
3851 if (skyhawk_chip(adapter
) && be_physfn(adapter
) &&
3852 !pci_vfs_assigned(pdev
)) {
3853 num_vf_qs
= be_calculate_vf_qs(adapter
,
3854 pci_sriov_get_totalvfs(pdev
));
3855 be_cmd_set_sriov_config(adapter
, adapter
->pool_res
,
3856 pci_sriov_get_totalvfs(pdev
),
3860 #ifdef CONFIG_BE2NET_VXLAN
3861 be_disable_vxlan_offloads(adapter
);
3863 kfree(adapter
->pmac_id
);
3864 adapter
->pmac_id
= NULL
;
3866 be_cmd_if_destroy(adapter
, adapter
->if_handle
, 0);
3868 be_clear_queues(adapter
);
3870 be_msix_disable(adapter
);
3871 adapter
->flags
&= ~BE_FLAGS_SETUP_DONE
;
3875 static int be_vfs_if_create(struct be_adapter
*adapter
)
3877 struct be_resources res
= {0};
3878 u32 cap_flags
, en_flags
, vf
;
3879 struct be_vf_cfg
*vf_cfg
;
3882 /* If a FW profile exists, then cap_flags are updated */
3883 cap_flags
= BE_VF_IF_EN_FLAGS
;
3885 for_all_vfs(adapter
, vf_cfg
, vf
) {
3886 if (!BE3_chip(adapter
)) {
3887 status
= be_cmd_get_profile_config(adapter
, &res
,
3891 cap_flags
= res
.if_cap_flags
;
3892 /* Prevent VFs from enabling VLAN promiscuous
3895 cap_flags
&= ~BE_IF_FLAGS_VLAN_PROMISCUOUS
;
3899 /* PF should enable IF flags during proxy if_create call */
3900 en_flags
= cap_flags
& BE_VF_IF_EN_FLAGS
;
3901 status
= be_cmd_if_create(adapter
, cap_flags
, en_flags
,
3902 &vf_cfg
->if_handle
, vf
+ 1);
3910 static int be_vf_setup_init(struct be_adapter
*adapter
)
3912 struct be_vf_cfg
*vf_cfg
;
3915 adapter
->vf_cfg
= kcalloc(adapter
->num_vfs
, sizeof(*vf_cfg
),
3917 if (!adapter
->vf_cfg
)
3920 for_all_vfs(adapter
, vf_cfg
, vf
) {
3921 vf_cfg
->if_handle
= -1;
3922 vf_cfg
->pmac_id
= -1;
3927 static int be_vf_setup(struct be_adapter
*adapter
)
3929 struct device
*dev
= &adapter
->pdev
->dev
;
3930 struct be_vf_cfg
*vf_cfg
;
3931 int status
, old_vfs
, vf
;
3934 old_vfs
= pci_num_vf(adapter
->pdev
);
3936 status
= be_vf_setup_init(adapter
);
3941 for_all_vfs(adapter
, vf_cfg
, vf
) {
3942 status
= be_cmd_get_if_id(adapter
, vf_cfg
, vf
);
3947 status
= be_vfs_mac_query(adapter
);
3951 status
= be_vfs_if_create(adapter
);
3955 status
= be_vf_eth_addr_config(adapter
);
3960 for_all_vfs(adapter
, vf_cfg
, vf
) {
3961 /* Allow VFs to programs MAC/VLAN filters */
3962 status
= be_cmd_get_fn_privileges(adapter
, &vf_cfg
->privileges
,
3964 if (!status
&& !(vf_cfg
->privileges
& BE_PRIV_FILTMGMT
)) {
3965 status
= be_cmd_set_fn_privileges(adapter
,
3966 vf_cfg
->privileges
|
3970 vf_cfg
->privileges
|= BE_PRIV_FILTMGMT
;
3971 dev_info(dev
, "VF%d has FILTMGMT privilege\n",
3976 /* Allow full available bandwidth */
3978 be_cmd_config_qos(adapter
, 0, 0, vf
+ 1);
3980 status
= be_cmd_get_hsw_config(adapter
, NULL
, vf
+ 1,
3981 vf_cfg
->if_handle
, NULL
,
3984 vf_cfg
->spoofchk
= spoofchk
;
3987 be_cmd_enable_vf(adapter
, vf
+ 1);
3988 be_cmd_set_logical_link_config(adapter
,
3989 IFLA_VF_LINK_STATE_AUTO
,
3995 status
= pci_enable_sriov(adapter
->pdev
, adapter
->num_vfs
);
3997 dev_err(dev
, "SRIOV enable failed\n");
3998 adapter
->num_vfs
= 0;
4003 adapter
->flags
|= BE_FLAGS_SRIOV_ENABLED
;
4006 dev_err(dev
, "VF setup failed\n");
4007 be_vf_clear(adapter
);
4011 /* Converting function_mode bits on BE3 to SH mc_type enums */
4013 static u8
be_convert_mc_type(u32 function_mode
)
4015 if (function_mode
& VNIC_MODE
&& function_mode
& QNQ_MODE
)
4017 else if (function_mode
& QNQ_MODE
)
4019 else if (function_mode
& VNIC_MODE
)
4021 else if (function_mode
& UMC_ENABLED
)
4027 /* On BE2/BE3 FW does not suggest the supported limits */
4028 static void BEx_get_resources(struct be_adapter
*adapter
,
4029 struct be_resources
*res
)
4031 bool use_sriov
= adapter
->num_vfs
? 1 : 0;
4033 if (be_physfn(adapter
))
4034 res
->max_uc_mac
= BE_UC_PMAC_COUNT
;
4036 res
->max_uc_mac
= BE_VF_UC_PMAC_COUNT
;
4038 adapter
->mc_type
= be_convert_mc_type(adapter
->function_mode
);
4040 if (be_is_mc(adapter
)) {
4041 /* Assuming that there are 4 channels per port,
4042 * when multi-channel is enabled
4044 if (be_is_qnq_mode(adapter
))
4045 res
->max_vlans
= BE_NUM_VLANS_SUPPORTED
/8;
4047 /* In a non-qnq multichannel mode, the pvid
4048 * takes up one vlan entry
4050 res
->max_vlans
= (BE_NUM_VLANS_SUPPORTED
/ 4) - 1;
4052 res
->max_vlans
= BE_NUM_VLANS_SUPPORTED
;
4055 res
->max_mcast_mac
= BE_MAX_MC
;
4057 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
4058 * 2) Create multiple TX rings on a BE3-R multi-channel interface
4059 * *only* if it is RSS-capable.
4061 if (BE2_chip(adapter
) || use_sriov
|| (adapter
->port_num
> 1) ||
4062 be_virtfn(adapter
) ||
4063 (be_is_mc(adapter
) &&
4064 !(adapter
->function_caps
& BE_FUNCTION_CAPS_RSS
))) {
4066 } else if (adapter
->function_caps
& BE_FUNCTION_CAPS_SUPER_NIC
) {
4067 struct be_resources super_nic_res
= {0};
4069 /* On a SuperNIC profile, the driver needs to use the
4070 * GET_PROFILE_CONFIG cmd to query the per-function TXQ limits
4072 be_cmd_get_profile_config(adapter
, &super_nic_res
,
4073 RESOURCE_LIMITS
, 0);
4074 /* Some old versions of BE3 FW don't report max_tx_qs value */
4075 res
->max_tx_qs
= super_nic_res
.max_tx_qs
? : BE3_MAX_TX_QS
;
4077 res
->max_tx_qs
= BE3_MAX_TX_QS
;
4080 if ((adapter
->function_caps
& BE_FUNCTION_CAPS_RSS
) &&
4081 !use_sriov
&& be_physfn(adapter
))
4082 res
->max_rss_qs
= (adapter
->be3_native
) ?
4083 BE3_MAX_RSS_QS
: BE2_MAX_RSS_QS
;
4084 res
->max_rx_qs
= res
->max_rss_qs
+ 1;
4086 if (be_physfn(adapter
))
4087 res
->max_evt_qs
= (be_max_vfs(adapter
) > 0) ?
4088 BE3_SRIOV_MAX_EVT_QS
: BE3_MAX_EVT_QS
;
4090 res
->max_evt_qs
= 1;
4092 res
->if_cap_flags
= BE_IF_CAP_FLAGS_WANT
;
4093 res
->if_cap_flags
&= ~BE_IF_FLAGS_DEFQ_RSS
;
4094 if (!(adapter
->function_caps
& BE_FUNCTION_CAPS_RSS
))
4095 res
->if_cap_flags
&= ~BE_IF_FLAGS_RSS
;
4098 static void be_setup_init(struct be_adapter
*adapter
)
4100 adapter
->vlan_prio_bmap
= 0xff;
4101 adapter
->phy
.link_speed
= -1;
4102 adapter
->if_handle
= -1;
4103 adapter
->be3_native
= false;
4104 adapter
->if_flags
= 0;
4105 adapter
->phy_state
= BE_UNKNOWN_PHY_STATE
;
4106 if (be_physfn(adapter
))
4107 adapter
->cmd_privileges
= MAX_PRIVILEGES
;
4109 adapter
->cmd_privileges
= MIN_PRIVILEGES
;
4112 static int be_get_sriov_config(struct be_adapter
*adapter
)
4114 struct be_resources res
= {0};
4115 int max_vfs
, old_vfs
;
4117 be_cmd_get_profile_config(adapter
, &res
, RESOURCE_LIMITS
, 0);
4119 /* Some old versions of BE3 FW don't report max_vfs value */
4120 if (BE3_chip(adapter
) && !res
.max_vfs
) {
4121 max_vfs
= pci_sriov_get_totalvfs(adapter
->pdev
);
4122 res
.max_vfs
= max_vfs
> 0 ? min(MAX_VFS
, max_vfs
) : 0;
4125 adapter
->pool_res
= res
;
4127 /* If during previous unload of the driver, the VFs were not disabled,
4128 * then we cannot rely on the PF POOL limits for the TotalVFs value.
4129 * Instead use the TotalVFs value stored in the pci-dev struct.
4131 old_vfs
= pci_num_vf(adapter
->pdev
);
4133 dev_info(&adapter
->pdev
->dev
, "%d VFs are already enabled\n",
4136 adapter
->pool_res
.max_vfs
=
4137 pci_sriov_get_totalvfs(adapter
->pdev
);
4138 adapter
->num_vfs
= old_vfs
;
4144 static void be_alloc_sriov_res(struct be_adapter
*adapter
)
4146 int old_vfs
= pci_num_vf(adapter
->pdev
);
4150 be_get_sriov_config(adapter
);
4153 pci_sriov_set_totalvfs(adapter
->pdev
, be_max_vfs(adapter
));
4155 /* When the HW is in SRIOV capable configuration, the PF-pool
4156 * resources are given to PF during driver load, if there are no
4157 * old VFs. This facility is not available in BE3 FW.
4158 * Also, this is done by FW in Lancer chip.
4160 if (skyhawk_chip(adapter
) && be_max_vfs(adapter
) && !old_vfs
) {
4161 num_vf_qs
= be_calculate_vf_qs(adapter
, 0);
4162 status
= be_cmd_set_sriov_config(adapter
, adapter
->pool_res
, 0,
4165 dev_err(&adapter
->pdev
->dev
,
4166 "Failed to optimize SRIOV resources\n");
4170 static int be_get_resources(struct be_adapter
*adapter
)
4172 struct device
*dev
= &adapter
->pdev
->dev
;
4173 struct be_resources res
= {0};
4176 if (BEx_chip(adapter
)) {
4177 BEx_get_resources(adapter
, &res
);
4181 /* For Lancer, SH etc read per-function resource limits from FW.
4182 * GET_FUNC_CONFIG returns per function guaranteed limits.
4183 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
4185 if (!BEx_chip(adapter
)) {
4186 status
= be_cmd_get_func_config(adapter
, &res
);
4190 /* If a deafault RXQ must be created, we'll use up one RSSQ*/
4191 if (res
.max_rss_qs
&& res
.max_rss_qs
== res
.max_rx_qs
&&
4192 !(res
.if_cap_flags
& BE_IF_FLAGS_DEFQ_RSS
))
4193 res
.max_rss_qs
-= 1;
4195 /* If RoCE may be enabled stash away half the EQs for RoCE */
4196 if (be_roce_supported(adapter
))
4197 res
.max_evt_qs
/= 2;
4201 /* If FW supports RSS default queue, then skip creating non-RSS
4202 * queue for non-IP traffic.
4204 adapter
->need_def_rxq
= (be_if_cap_flags(adapter
) &
4205 BE_IF_FLAGS_DEFQ_RSS
) ? 0 : 1;
4207 dev_info(dev
, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
4208 be_max_txqs(adapter
), be_max_rxqs(adapter
),
4209 be_max_rss(adapter
), be_max_eqs(adapter
),
4210 be_max_vfs(adapter
));
4211 dev_info(dev
, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
4212 be_max_uc(adapter
), be_max_mc(adapter
),
4213 be_max_vlans(adapter
));
4215 /* Sanitize cfg_num_qs based on HW and platform limits */
4216 adapter
->cfg_num_qs
= min_t(u16
, netif_get_num_default_rss_queues(),
4217 be_max_qs(adapter
));
4221 static int be_get_config(struct be_adapter
*adapter
)
4226 status
= be_cmd_get_cntl_attributes(adapter
);
4230 status
= be_cmd_query_fw_cfg(adapter
);
4234 if (!lancer_chip(adapter
) && be_physfn(adapter
))
4235 be_cmd_get_fat_dump_len(adapter
, &adapter
->fat_dump_len
);
4237 if (BEx_chip(adapter
)) {
4238 level
= be_cmd_get_fw_log_level(adapter
);
4239 adapter
->msg_enable
=
4240 level
<= FW_LOG_LEVEL_DEFAULT
? NETIF_MSG_HW
: 0;
4243 be_cmd_get_acpi_wol_cap(adapter
);
4245 be_cmd_query_port_name(adapter
);
4247 if (be_physfn(adapter
)) {
4248 status
= be_cmd_get_active_profile(adapter
, &profile_id
);
4250 dev_info(&adapter
->pdev
->dev
,
4251 "Using profile 0x%x\n", profile_id
);
4254 status
= be_get_resources(adapter
);
4258 adapter
->pmac_id
= kcalloc(be_max_uc(adapter
),
4259 sizeof(*adapter
->pmac_id
), GFP_KERNEL
);
4260 if (!adapter
->pmac_id
)
4266 static int be_mac_setup(struct be_adapter
*adapter
)
4271 if (is_zero_ether_addr(adapter
->netdev
->dev_addr
)) {
4272 status
= be_cmd_get_perm_mac(adapter
, mac
);
4276 memcpy(adapter
->netdev
->dev_addr
, mac
, ETH_ALEN
);
4277 memcpy(adapter
->netdev
->perm_addr
, mac
, ETH_ALEN
);
4283 static void be_schedule_worker(struct be_adapter
*adapter
)
4285 schedule_delayed_work(&adapter
->work
, msecs_to_jiffies(1000));
4286 adapter
->flags
|= BE_FLAGS_WORKER_SCHEDULED
;
4289 static void be_schedule_err_detection(struct be_adapter
*adapter
, u32 delay
)
4291 schedule_delayed_work(&adapter
->be_err_detection_work
,
4292 msecs_to_jiffies(delay
));
4293 adapter
->flags
|= BE_FLAGS_ERR_DETECTION_SCHEDULED
;
4296 static int be_setup_queues(struct be_adapter
*adapter
)
4298 struct net_device
*netdev
= adapter
->netdev
;
4301 status
= be_evt_queues_create(adapter
);
4305 status
= be_tx_qs_create(adapter
);
4309 status
= be_rx_cqs_create(adapter
);
4313 status
= be_mcc_queues_create(adapter
);
4317 status
= netif_set_real_num_rx_queues(netdev
, adapter
->num_rx_qs
);
4321 status
= netif_set_real_num_tx_queues(netdev
, adapter
->num_tx_qs
);
4327 dev_err(&adapter
->pdev
->dev
, "queue_setup failed\n");
4331 static int be_if_create(struct be_adapter
*adapter
)
4333 u32 en_flags
= BE_IF_FLAGS_RSS
| BE_IF_FLAGS_DEFQ_RSS
;
4334 u32 cap_flags
= be_if_cap_flags(adapter
);
4337 if (adapter
->cfg_num_qs
== 1)
4338 cap_flags
&= ~(BE_IF_FLAGS_DEFQ_RSS
| BE_IF_FLAGS_RSS
);
4340 en_flags
&= cap_flags
;
4341 /* will enable all the needed filter flags in be_open() */
4342 status
= be_cmd_if_create(adapter
, be_if_cap_flags(adapter
), en_flags
,
4343 &adapter
->if_handle
, 0);
4348 int be_update_queues(struct be_adapter
*adapter
)
4350 struct net_device
*netdev
= adapter
->netdev
;
4353 if (netif_running(netdev
))
4356 be_cancel_worker(adapter
);
4358 /* If any vectors have been shared with RoCE we cannot re-program
4361 if (!adapter
->num_msix_roce_vec
)
4362 be_msix_disable(adapter
);
4364 be_clear_queues(adapter
);
4365 status
= be_cmd_if_destroy(adapter
, adapter
->if_handle
, 0);
4369 if (!msix_enabled(adapter
)) {
4370 status
= be_msix_enable(adapter
);
4375 status
= be_if_create(adapter
);
4379 status
= be_setup_queues(adapter
);
4383 be_schedule_worker(adapter
);
4385 if (netif_running(netdev
))
4386 status
= be_open(netdev
);
4391 static inline int fw_major_num(const char *fw_ver
)
4393 int fw_major
= 0, i
;
4395 i
= sscanf(fw_ver
, "%d.", &fw_major
);
4402 /* If any VFs are already enabled don't FLR the PF */
4403 static bool be_reset_required(struct be_adapter
*adapter
)
4405 return pci_num_vf(adapter
->pdev
) ? false : true;
4408 /* Wait for the FW to be ready and perform the required initialization */
4409 static int be_func_init(struct be_adapter
*adapter
)
4413 status
= be_fw_wait_ready(adapter
);
4417 if (be_reset_required(adapter
)) {
4418 status
= be_cmd_reset_function(adapter
);
4422 /* Wait for interrupts to quiesce after an FLR */
4425 /* We can clear all errors when function reset succeeds */
4426 be_clear_error(adapter
, BE_CLEAR_ALL
);
4429 /* Tell FW we're ready to fire cmds */
4430 status
= be_cmd_fw_init(adapter
);
4434 /* Allow interrupts for other ULPs running on NIC function */
4435 be_intr_set(adapter
, true);
4440 static int be_setup(struct be_adapter
*adapter
)
4442 struct device
*dev
= &adapter
->pdev
->dev
;
4445 status
= be_func_init(adapter
);
4449 be_setup_init(adapter
);
4451 if (!lancer_chip(adapter
))
4452 be_cmd_req_native_mode(adapter
);
4454 /* invoke this cmd first to get pf_num and vf_num which are needed
4455 * for issuing profile related cmds
4457 if (!BEx_chip(adapter
)) {
4458 status
= be_cmd_get_func_config(adapter
, NULL
);
4463 if (!BE2_chip(adapter
) && be_physfn(adapter
))
4464 be_alloc_sriov_res(adapter
);
4466 status
= be_get_config(adapter
);
4470 status
= be_msix_enable(adapter
);
4474 /* will enable all the needed filter flags in be_open() */
4475 status
= be_if_create(adapter
);
4479 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
4481 status
= be_setup_queues(adapter
);
4486 be_cmd_get_fn_privileges(adapter
, &adapter
->cmd_privileges
, 0);
4488 status
= be_mac_setup(adapter
);
4492 be_cmd_get_fw_ver(adapter
);
4493 dev_info(dev
, "FW version is %s\n", adapter
->fw_ver
);
4495 if (BE2_chip(adapter
) && fw_major_num(adapter
->fw_ver
) < 4) {
4496 dev_err(dev
, "Firmware on card is old(%s), IRQs may not work",
4498 dev_err(dev
, "Please upgrade firmware to version >= 4.0\n");
4501 status
= be_cmd_set_flow_control(adapter
, adapter
->tx_fc
,
4504 be_cmd_get_flow_control(adapter
, &adapter
->tx_fc
,
4507 dev_info(&adapter
->pdev
->dev
, "HW Flow control - TX:%d RX:%d\n",
4508 adapter
->tx_fc
, adapter
->rx_fc
);
4510 if (be_physfn(adapter
))
4511 be_cmd_set_logical_link_config(adapter
,
4512 IFLA_VF_LINK_STATE_AUTO
, 0);
4514 if (adapter
->num_vfs
)
4515 be_vf_setup(adapter
);
4517 status
= be_cmd_get_phy_info(adapter
);
4518 if (!status
&& be_pause_supported(adapter
))
4519 adapter
->phy
.fc_autoneg
= 1;
4521 be_schedule_worker(adapter
);
4522 adapter
->flags
|= BE_FLAGS_SETUP_DONE
;
4529 #ifdef CONFIG_NET_POLL_CONTROLLER
4530 static void be_netpoll(struct net_device
*netdev
)
4532 struct be_adapter
*adapter
= netdev_priv(netdev
);
4533 struct be_eq_obj
*eqo
;
4536 for_all_evt_queues(adapter
, eqo
, i
) {
4537 be_eq_notify(eqo
->adapter
, eqo
->q
.id
, false, true, 0, 0);
4538 napi_schedule(&eqo
->napi
);
4543 int be_load_fw(struct be_adapter
*adapter
, u8
*fw_file
)
4545 const struct firmware
*fw
;
4548 if (!netif_running(adapter
->netdev
)) {
4549 dev_err(&adapter
->pdev
->dev
,
4550 "Firmware load not allowed (interface is down)\n");
4554 status
= request_firmware(&fw
, fw_file
, &adapter
->pdev
->dev
);
4558 dev_info(&adapter
->pdev
->dev
, "Flashing firmware file %s\n", fw_file
);
4560 if (lancer_chip(adapter
))
4561 status
= lancer_fw_download(adapter
, fw
);
4563 status
= be_fw_download(adapter
, fw
);
4566 be_cmd_get_fw_ver(adapter
);
4569 release_firmware(fw
);
4573 static int be_ndo_bridge_setlink(struct net_device
*dev
, struct nlmsghdr
*nlh
,
4576 struct be_adapter
*adapter
= netdev_priv(dev
);
4577 struct nlattr
*attr
, *br_spec
;
4582 if (!sriov_enabled(adapter
))
4585 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
4589 nla_for_each_nested(attr
, br_spec
, rem
) {
4590 if (nla_type(attr
) != IFLA_BRIDGE_MODE
)
4593 if (nla_len(attr
) < sizeof(mode
))
4596 mode
= nla_get_u16(attr
);
4597 if (BE3_chip(adapter
) && mode
== BRIDGE_MODE_VEPA
)
4600 if (mode
!= BRIDGE_MODE_VEPA
&& mode
!= BRIDGE_MODE_VEB
)
4603 status
= be_cmd_set_hsw_config(adapter
, 0, 0,
4605 mode
== BRIDGE_MODE_VEPA
?
4606 PORT_FWD_TYPE_VEPA
:
4607 PORT_FWD_TYPE_VEB
, 0);
4611 dev_info(&adapter
->pdev
->dev
, "enabled switch mode: %s\n",
4612 mode
== BRIDGE_MODE_VEPA
? "VEPA" : "VEB");
4617 dev_err(&adapter
->pdev
->dev
, "Failed to set switch mode %s\n",
4618 mode
== BRIDGE_MODE_VEPA
? "VEPA" : "VEB");
4623 static int be_ndo_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
4624 struct net_device
*dev
, u32 filter_mask
,
4627 struct be_adapter
*adapter
= netdev_priv(dev
);
4631 /* BE and Lancer chips support VEB mode only */
4632 if (BEx_chip(adapter
) || lancer_chip(adapter
)) {
4633 /* VEB is disabled in non-SR-IOV profiles on BE3/Lancer */
4634 if (!pci_sriov_get_totalvfs(adapter
->pdev
))
4636 hsw_mode
= PORT_FWD_TYPE_VEB
;
4638 status
= be_cmd_get_hsw_config(adapter
, NULL
, 0,
4639 adapter
->if_handle
, &hsw_mode
,
4644 if (hsw_mode
== PORT_FWD_TYPE_PASSTHRU
)
4648 return ndo_dflt_bridge_getlink(skb
, pid
, seq
, dev
,
4649 hsw_mode
== PORT_FWD_TYPE_VEPA
?
4650 BRIDGE_MODE_VEPA
: BRIDGE_MODE_VEB
,
4651 0, 0, nlflags
, filter_mask
, NULL
);
4654 #ifdef CONFIG_BE2NET_VXLAN
4655 /* VxLAN offload Notes:
4657 * The stack defines tunnel offload flags (hw_enc_features) for IP and doesn't
4658 * distinguish various types of transports (VxLAN, GRE, NVGRE ..). So, offload
4659 * is expected to work across all types of IP tunnels once exported. Skyhawk
4660 * supports offloads for either VxLAN or NVGRE, exclusively. So we export VxLAN
4661 * offloads in hw_enc_features only when a VxLAN port is added. If other (non
4662 * VxLAN) tunnels are configured while VxLAN offloads are enabled, offloads for
4663 * those other tunnels are unexported on the fly through ndo_features_check().
4665 * Skyhawk supports VxLAN offloads only for one UDP dport. So, if the stack
4666 * adds more than one port, disable offloads and don't re-enable them again
4667 * until after all the tunnels are removed.
4669 static void be_add_vxlan_port(struct net_device
*netdev
, sa_family_t sa_family
,
4672 struct be_adapter
*adapter
= netdev_priv(netdev
);
4673 struct device
*dev
= &adapter
->pdev
->dev
;
4676 if (lancer_chip(adapter
) || BEx_chip(adapter
) || be_is_mc(adapter
))
4679 if (adapter
->vxlan_port
== port
&& adapter
->vxlan_port_count
) {
4680 adapter
->vxlan_port_aliases
++;
4684 if (adapter
->flags
& BE_FLAGS_VXLAN_OFFLOADS
) {
4686 "Only one UDP port supported for VxLAN offloads\n");
4687 dev_info(dev
, "Disabling VxLAN offloads\n");
4688 adapter
->vxlan_port_count
++;
4692 if (adapter
->vxlan_port_count
++ >= 1)
4695 status
= be_cmd_manage_iface(adapter
, adapter
->if_handle
,
4696 OP_CONVERT_NORMAL_TO_TUNNEL
);
4698 dev_warn(dev
, "Failed to convert normal interface to tunnel\n");
4702 status
= be_cmd_set_vxlan_port(adapter
, port
);
4704 dev_warn(dev
, "Failed to add VxLAN port\n");
4707 adapter
->flags
|= BE_FLAGS_VXLAN_OFFLOADS
;
4708 adapter
->vxlan_port
= port
;
4710 netdev
->hw_enc_features
|= NETIF_F_IP_CSUM
| NETIF_F_IPV6_CSUM
|
4711 NETIF_F_TSO
| NETIF_F_TSO6
|
4712 NETIF_F_GSO_UDP_TUNNEL
;
4713 netdev
->hw_features
|= NETIF_F_GSO_UDP_TUNNEL
;
4714 netdev
->features
|= NETIF_F_GSO_UDP_TUNNEL
;
4716 dev_info(dev
, "Enabled VxLAN offloads for UDP port %d\n",
4720 be_disable_vxlan_offloads(adapter
);
4723 static void be_del_vxlan_port(struct net_device
*netdev
, sa_family_t sa_family
,
4726 struct be_adapter
*adapter
= netdev_priv(netdev
);
4728 if (lancer_chip(adapter
) || BEx_chip(adapter
) || be_is_mc(adapter
))
4731 if (adapter
->vxlan_port
!= port
)
4734 if (adapter
->vxlan_port_aliases
) {
4735 adapter
->vxlan_port_aliases
--;
4739 be_disable_vxlan_offloads(adapter
);
4741 dev_info(&adapter
->pdev
->dev
,
4742 "Disabled VxLAN offloads for UDP port %d\n",
4745 adapter
->vxlan_port_count
--;
4748 static netdev_features_t
be_features_check(struct sk_buff
*skb
,
4749 struct net_device
*dev
,
4750 netdev_features_t features
)
4752 struct be_adapter
*adapter
= netdev_priv(dev
);
4755 /* The code below restricts offload features for some tunneled packets.
4756 * Offload features for normal (non tunnel) packets are unchanged.
4758 if (!skb
->encapsulation
||
4759 !(adapter
->flags
& BE_FLAGS_VXLAN_OFFLOADS
))
4762 /* It's an encapsulated packet and VxLAN offloads are enabled. We
4763 * should disable tunnel offload features if it's not a VxLAN packet,
4764 * as tunnel offloads have been enabled only for VxLAN. This is done to
4765 * allow other tunneled traffic like GRE work fine while VxLAN
4766 * offloads are configured in Skyhawk-R.
4768 switch (vlan_get_protocol(skb
)) {
4769 case htons(ETH_P_IP
):
4770 l4_hdr
= ip_hdr(skb
)->protocol
;
4772 case htons(ETH_P_IPV6
):
4773 l4_hdr
= ipv6_hdr(skb
)->nexthdr
;
4779 if (l4_hdr
!= IPPROTO_UDP
||
4780 skb
->inner_protocol_type
!= ENCAP_TYPE_ETHER
||
4781 skb
->inner_protocol
!= htons(ETH_P_TEB
) ||
4782 skb_inner_mac_header(skb
) - skb_transport_header(skb
) !=
4783 sizeof(struct udphdr
) + sizeof(struct vxlanhdr
))
4784 return features
& ~(NETIF_F_CSUM_MASK
| NETIF_F_GSO_MASK
);
4790 static int be_get_phys_port_id(struct net_device
*dev
,
4791 struct netdev_phys_item_id
*ppid
)
4793 int i
, id_len
= CNTL_SERIAL_NUM_WORDS
* CNTL_SERIAL_NUM_WORD_SZ
+ 1;
4794 struct be_adapter
*adapter
= netdev_priv(dev
);
4797 if (MAX_PHYS_ITEM_ID_LEN
< id_len
)
4800 ppid
->id
[0] = adapter
->hba_port_num
+ 1;
4802 for (i
= CNTL_SERIAL_NUM_WORDS
- 1; i
>= 0;
4803 i
--, id
+= CNTL_SERIAL_NUM_WORD_SZ
)
4804 memcpy(id
, &adapter
->serial_num
[i
], CNTL_SERIAL_NUM_WORD_SZ
);
4806 ppid
->id_len
= id_len
;
4811 static const struct net_device_ops be_netdev_ops
= {
4812 .ndo_open
= be_open
,
4813 .ndo_stop
= be_close
,
4814 .ndo_start_xmit
= be_xmit
,
4815 .ndo_set_rx_mode
= be_set_rx_mode
,
4816 .ndo_set_mac_address
= be_mac_addr_set
,
4817 .ndo_change_mtu
= be_change_mtu
,
4818 .ndo_get_stats64
= be_get_stats64
,
4819 .ndo_validate_addr
= eth_validate_addr
,
4820 .ndo_vlan_rx_add_vid
= be_vlan_add_vid
,
4821 .ndo_vlan_rx_kill_vid
= be_vlan_rem_vid
,
4822 .ndo_set_vf_mac
= be_set_vf_mac
,
4823 .ndo_set_vf_vlan
= be_set_vf_vlan
,
4824 .ndo_set_vf_rate
= be_set_vf_tx_rate
,
4825 .ndo_get_vf_config
= be_get_vf_config
,
4826 .ndo_set_vf_link_state
= be_set_vf_link_state
,
4827 .ndo_set_vf_spoofchk
= be_set_vf_spoofchk
,
4828 #ifdef CONFIG_NET_POLL_CONTROLLER
4829 .ndo_poll_controller
= be_netpoll
,
4831 .ndo_bridge_setlink
= be_ndo_bridge_setlink
,
4832 .ndo_bridge_getlink
= be_ndo_bridge_getlink
,
4833 #ifdef CONFIG_NET_RX_BUSY_POLL
4834 .ndo_busy_poll
= be_busy_poll
,
4836 #ifdef CONFIG_BE2NET_VXLAN
4837 .ndo_add_vxlan_port
= be_add_vxlan_port
,
4838 .ndo_del_vxlan_port
= be_del_vxlan_port
,
4839 .ndo_features_check
= be_features_check
,
4841 .ndo_get_phys_port_id
= be_get_phys_port_id
,
4844 static void be_netdev_init(struct net_device
*netdev
)
4846 struct be_adapter
*adapter
= netdev_priv(netdev
);
4848 netdev
->hw_features
|= NETIF_F_SG
| NETIF_F_TSO
| NETIF_F_TSO6
|
4849 NETIF_F_IP_CSUM
| NETIF_F_IPV6_CSUM
| NETIF_F_RXCSUM
|
4850 NETIF_F_HW_VLAN_CTAG_TX
;
4851 if ((be_if_cap_flags(adapter
) & BE_IF_FLAGS_RSS
))
4852 netdev
->hw_features
|= NETIF_F_RXHASH
;
4854 netdev
->features
|= netdev
->hw_features
|
4855 NETIF_F_HW_VLAN_CTAG_RX
| NETIF_F_HW_VLAN_CTAG_FILTER
;
4857 netdev
->vlan_features
|= NETIF_F_SG
| NETIF_F_TSO
| NETIF_F_TSO6
|
4858 NETIF_F_IP_CSUM
| NETIF_F_IPV6_CSUM
;
4860 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
4862 netdev
->flags
|= IFF_MULTICAST
;
4864 netif_set_gso_max_size(netdev
, BE_MAX_GSO_SIZE
- ETH_HLEN
);
4866 netdev
->netdev_ops
= &be_netdev_ops
;
4868 netdev
->ethtool_ops
= &be_ethtool_ops
;
4871 static void be_cleanup(struct be_adapter
*adapter
)
4873 struct net_device
*netdev
= adapter
->netdev
;
4876 netif_device_detach(netdev
);
4877 if (netif_running(netdev
))
4884 static int be_resume(struct be_adapter
*adapter
)
4886 struct net_device
*netdev
= adapter
->netdev
;
4889 status
= be_setup(adapter
);
4893 if (netif_running(netdev
)) {
4894 status
= be_open(netdev
);
4899 netif_device_attach(netdev
);
4904 static int be_err_recover(struct be_adapter
*adapter
)
4908 /* Error recovery is supported only Lancer as of now */
4909 if (!lancer_chip(adapter
))
4912 /* Wait for adapter to reach quiescent state before
4915 status
= be_fw_wait_ready(adapter
);
4919 be_cleanup(adapter
);
4921 status
= be_resume(adapter
);
4930 static void be_err_detection_task(struct work_struct
*work
)
4932 struct be_adapter
*adapter
=
4933 container_of(work
, struct be_adapter
,
4934 be_err_detection_work
.work
);
4935 struct device
*dev
= &adapter
->pdev
->dev
;
4936 int recovery_status
;
4937 int delay
= ERR_DETECTION_DELAY
;
4939 be_detect_error(adapter
);
4941 if (be_check_error(adapter
, BE_ERROR_HW
))
4942 recovery_status
= be_err_recover(adapter
);
4944 goto reschedule_task
;
4946 if (!recovery_status
) {
4947 adapter
->recovery_retries
= 0;
4948 dev_info(dev
, "Adapter recovery successful\n");
4949 goto reschedule_task
;
4950 } else if (be_virtfn(adapter
)) {
4951 /* For VFs, check if PF have allocated resources
4954 dev_err(dev
, "Re-trying adapter recovery\n");
4955 goto reschedule_task
;
4956 } else if (adapter
->recovery_retries
++ <
4957 MAX_ERR_RECOVERY_RETRY_COUNT
) {
4958 /* In case of another error during recovery, it takes 30 sec
4959 * for adapter to come out of error. Retry error recovery after
4960 * this time interval.
4962 dev_err(&adapter
->pdev
->dev
, "Re-trying adapter recovery\n");
4963 delay
= ERR_RECOVERY_RETRY_DELAY
;
4964 goto reschedule_task
;
4966 dev_err(dev
, "Adapter recovery failed\n");
4971 be_schedule_err_detection(adapter
, delay
);
4974 static void be_log_sfp_info(struct be_adapter
*adapter
)
4978 status
= be_cmd_query_sfp_info(adapter
);
4980 dev_err(&adapter
->pdev
->dev
,
4981 "Port %c: %s Vendor: %s part no: %s",
4983 be_misconfig_evt_port_state
[adapter
->phy_state
],
4984 adapter
->phy
.vendor_name
,
4985 adapter
->phy
.vendor_pn
);
4987 adapter
->flags
&= ~BE_FLAGS_PHY_MISCONFIGURED
;
4990 static void be_worker(struct work_struct
*work
)
4992 struct be_adapter
*adapter
=
4993 container_of(work
, struct be_adapter
, work
.work
);
4994 struct be_rx_obj
*rxo
;
4997 /* when interrupts are not yet enabled, just reap any pending
5000 if (!netif_running(adapter
->netdev
)) {
5002 be_process_mcc(adapter
);
5007 if (!adapter
->stats_cmd_sent
) {
5008 if (lancer_chip(adapter
))
5009 lancer_cmd_get_pport_stats(adapter
,
5010 &adapter
->stats_cmd
);
5012 be_cmd_get_stats(adapter
, &adapter
->stats_cmd
);
5015 if (be_physfn(adapter
) &&
5016 MODULO(adapter
->work_counter
, adapter
->be_get_temp_freq
) == 0)
5017 be_cmd_get_die_temperature(adapter
);
5019 for_all_rx_queues(adapter
, rxo
, i
) {
5020 /* Replenish RX-queues starved due to memory
5021 * allocation failures.
5023 if (rxo
->rx_post_starved
)
5024 be_post_rx_frags(rxo
, GFP_KERNEL
, MAX_RX_POST
);
5027 /* EQ-delay update for Skyhawk is done while notifying EQ */
5028 if (!skyhawk_chip(adapter
))
5029 be_eqd_update(adapter
, false);
5031 if (adapter
->flags
& BE_FLAGS_PHY_MISCONFIGURED
)
5032 be_log_sfp_info(adapter
);
5035 adapter
->work_counter
++;
5036 schedule_delayed_work(&adapter
->work
, msecs_to_jiffies(1000));
5039 static void be_unmap_pci_bars(struct be_adapter
*adapter
)
5042 pci_iounmap(adapter
->pdev
, adapter
->csr
);
5044 pci_iounmap(adapter
->pdev
, adapter
->db
);
5045 if (adapter
->pcicfg
&& adapter
->pcicfg_mapped
)
5046 pci_iounmap(adapter
->pdev
, adapter
->pcicfg
);
5049 static int db_bar(struct be_adapter
*adapter
)
5051 if (lancer_chip(adapter
) || be_virtfn(adapter
))
5057 static int be_roce_map_pci_bars(struct be_adapter
*adapter
)
5059 if (skyhawk_chip(adapter
)) {
5060 adapter
->roce_db
.size
= 4096;
5061 adapter
->roce_db
.io_addr
= pci_resource_start(adapter
->pdev
,
5063 adapter
->roce_db
.total_size
= pci_resource_len(adapter
->pdev
,
5069 static int be_map_pci_bars(struct be_adapter
*adapter
)
5071 struct pci_dev
*pdev
= adapter
->pdev
;
5075 pci_read_config_dword(adapter
->pdev
, SLI_INTF_REG_OFFSET
, &sli_intf
);
5076 adapter
->sli_family
= (sli_intf
& SLI_INTF_FAMILY_MASK
) >>
5077 SLI_INTF_FAMILY_SHIFT
;
5078 adapter
->virtfn
= (sli_intf
& SLI_INTF_FT_MASK
) ? 1 : 0;
5080 if (BEx_chip(adapter
) && be_physfn(adapter
)) {
5081 adapter
->csr
= pci_iomap(pdev
, 2, 0);
5086 addr
= pci_iomap(pdev
, db_bar(adapter
), 0);
5091 if (skyhawk_chip(adapter
) || BEx_chip(adapter
)) {
5092 if (be_physfn(adapter
)) {
5093 /* PCICFG is the 2nd BAR in BE2 */
5094 addr
= pci_iomap(pdev
, BE2_chip(adapter
) ? 1 : 0, 0);
5097 adapter
->pcicfg
= addr
;
5098 adapter
->pcicfg_mapped
= true;
5100 adapter
->pcicfg
= adapter
->db
+ SRIOV_VF_PCICFG_OFFSET
;
5101 adapter
->pcicfg_mapped
= false;
5105 be_roce_map_pci_bars(adapter
);
5109 dev_err(&pdev
->dev
, "Error in mapping PCI BARs\n");
5110 be_unmap_pci_bars(adapter
);
5114 static void be_drv_cleanup(struct be_adapter
*adapter
)
5116 struct be_dma_mem
*mem
= &adapter
->mbox_mem_alloced
;
5117 struct device
*dev
= &adapter
->pdev
->dev
;
5120 dma_free_coherent(dev
, mem
->size
, mem
->va
, mem
->dma
);
5122 mem
= &adapter
->rx_filter
;
5124 dma_free_coherent(dev
, mem
->size
, mem
->va
, mem
->dma
);
5126 mem
= &adapter
->stats_cmd
;
5128 dma_free_coherent(dev
, mem
->size
, mem
->va
, mem
->dma
);
5131 /* Allocate and initialize various fields in be_adapter struct */
5132 static int be_drv_init(struct be_adapter
*adapter
)
5134 struct be_dma_mem
*mbox_mem_alloc
= &adapter
->mbox_mem_alloced
;
5135 struct be_dma_mem
*mbox_mem_align
= &adapter
->mbox_mem
;
5136 struct be_dma_mem
*rx_filter
= &adapter
->rx_filter
;
5137 struct be_dma_mem
*stats_cmd
= &adapter
->stats_cmd
;
5138 struct device
*dev
= &adapter
->pdev
->dev
;
5141 mbox_mem_alloc
->size
= sizeof(struct be_mcc_mailbox
) + 16;
5142 mbox_mem_alloc
->va
= dma_zalloc_coherent(dev
, mbox_mem_alloc
->size
,
5143 &mbox_mem_alloc
->dma
,
5145 if (!mbox_mem_alloc
->va
)
5148 mbox_mem_align
->size
= sizeof(struct be_mcc_mailbox
);
5149 mbox_mem_align
->va
= PTR_ALIGN(mbox_mem_alloc
->va
, 16);
5150 mbox_mem_align
->dma
= PTR_ALIGN(mbox_mem_alloc
->dma
, 16);
5152 rx_filter
->size
= sizeof(struct be_cmd_req_rx_filter
);
5153 rx_filter
->va
= dma_zalloc_coherent(dev
, rx_filter
->size
,
5154 &rx_filter
->dma
, GFP_KERNEL
);
5155 if (!rx_filter
->va
) {
5160 if (lancer_chip(adapter
))
5161 stats_cmd
->size
= sizeof(struct lancer_cmd_req_pport_stats
);
5162 else if (BE2_chip(adapter
))
5163 stats_cmd
->size
= sizeof(struct be_cmd_req_get_stats_v0
);
5164 else if (BE3_chip(adapter
))
5165 stats_cmd
->size
= sizeof(struct be_cmd_req_get_stats_v1
);
5167 stats_cmd
->size
= sizeof(struct be_cmd_req_get_stats_v2
);
5168 stats_cmd
->va
= dma_zalloc_coherent(dev
, stats_cmd
->size
,
5169 &stats_cmd
->dma
, GFP_KERNEL
);
5170 if (!stats_cmd
->va
) {
5172 goto free_rx_filter
;
5175 mutex_init(&adapter
->mbox_lock
);
5176 spin_lock_init(&adapter
->mcc_lock
);
5177 spin_lock_init(&adapter
->mcc_cq_lock
);
5178 init_completion(&adapter
->et_cmd_compl
);
5180 pci_save_state(adapter
->pdev
);
5182 INIT_DELAYED_WORK(&adapter
->work
, be_worker
);
5183 INIT_DELAYED_WORK(&adapter
->be_err_detection_work
,
5184 be_err_detection_task
);
5186 adapter
->rx_fc
= true;
5187 adapter
->tx_fc
= true;
5189 /* Must be a power of 2 or else MODULO will BUG_ON */
5190 adapter
->be_get_temp_freq
= 64;
5195 dma_free_coherent(dev
, rx_filter
->size
, rx_filter
->va
, rx_filter
->dma
);
5197 dma_free_coherent(dev
, mbox_mem_alloc
->size
, mbox_mem_alloc
->va
,
5198 mbox_mem_alloc
->dma
);
5202 static void be_remove(struct pci_dev
*pdev
)
5204 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5209 be_roce_dev_remove(adapter
);
5210 be_intr_set(adapter
, false);
5212 be_cancel_err_detection(adapter
);
5214 unregister_netdev(adapter
->netdev
);
5218 /* tell fw we're done with firing cmds */
5219 be_cmd_fw_clean(adapter
);
5221 be_unmap_pci_bars(adapter
);
5222 be_drv_cleanup(adapter
);
5224 pci_disable_pcie_error_reporting(pdev
);
5226 pci_release_regions(pdev
);
5227 pci_disable_device(pdev
);
5229 free_netdev(adapter
->netdev
);
5232 static ssize_t
be_hwmon_show_temp(struct device
*dev
,
5233 struct device_attribute
*dev_attr
,
5236 struct be_adapter
*adapter
= dev_get_drvdata(dev
);
5238 /* Unit: millidegree Celsius */
5239 if (adapter
->hwmon_info
.be_on_die_temp
== BE_INVALID_DIE_TEMP
)
5242 return sprintf(buf
, "%u\n",
5243 adapter
->hwmon_info
.be_on_die_temp
* 1000);
5246 static SENSOR_DEVICE_ATTR(temp1_input
, S_IRUGO
,
5247 be_hwmon_show_temp
, NULL
, 1);
5249 static struct attribute
*be_hwmon_attrs
[] = {
5250 &sensor_dev_attr_temp1_input
.dev_attr
.attr
,
5254 ATTRIBUTE_GROUPS(be_hwmon
);
5256 static char *mc_name(struct be_adapter
*adapter
)
5258 char *str
= ""; /* default */
5260 switch (adapter
->mc_type
) {
5286 static inline char *func_name(struct be_adapter
*adapter
)
5288 return be_physfn(adapter
) ? "PF" : "VF";
5291 static inline char *nic_name(struct pci_dev
*pdev
)
5293 switch (pdev
->device
) {
5300 return OC_NAME_LANCER
;
5311 static int be_probe(struct pci_dev
*pdev
, const struct pci_device_id
*pdev_id
)
5313 struct be_adapter
*adapter
;
5314 struct net_device
*netdev
;
5317 dev_info(&pdev
->dev
, "%s version is %s\n", DRV_NAME
, DRV_VER
);
5319 status
= pci_enable_device(pdev
);
5323 status
= pci_request_regions(pdev
, DRV_NAME
);
5326 pci_set_master(pdev
);
5328 netdev
= alloc_etherdev_mqs(sizeof(*adapter
), MAX_TX_QS
, MAX_RX_QS
);
5333 adapter
= netdev_priv(netdev
);
5334 adapter
->pdev
= pdev
;
5335 pci_set_drvdata(pdev
, adapter
);
5336 adapter
->netdev
= netdev
;
5337 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
5339 status
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(64));
5341 netdev
->features
|= NETIF_F_HIGHDMA
;
5343 status
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(32));
5345 dev_err(&pdev
->dev
, "Could not set PCI DMA Mask\n");
5350 status
= pci_enable_pcie_error_reporting(pdev
);
5352 dev_info(&pdev
->dev
, "PCIe error reporting enabled\n");
5354 status
= be_map_pci_bars(adapter
);
5358 status
= be_drv_init(adapter
);
5362 status
= be_setup(adapter
);
5366 be_netdev_init(netdev
);
5367 status
= register_netdev(netdev
);
5371 be_roce_dev_add(adapter
);
5373 be_schedule_err_detection(adapter
, ERR_DETECTION_DELAY
);
5375 /* On Die temperature not supported for VF. */
5376 if (be_physfn(adapter
) && IS_ENABLED(CONFIG_BE2NET_HWMON
)) {
5377 adapter
->hwmon_info
.hwmon_dev
=
5378 devm_hwmon_device_register_with_groups(&pdev
->dev
,
5382 adapter
->hwmon_info
.be_on_die_temp
= BE_INVALID_DIE_TEMP
;
5385 dev_info(&pdev
->dev
, "%s: %s %s port %c\n", nic_name(pdev
),
5386 func_name(adapter
), mc_name(adapter
), adapter
->port_name
);
5393 be_drv_cleanup(adapter
);
5395 be_unmap_pci_bars(adapter
);
5397 free_netdev(netdev
);
5399 pci_release_regions(pdev
);
5401 pci_disable_device(pdev
);
5403 dev_err(&pdev
->dev
, "%s initialization failed\n", nic_name(pdev
));
5407 static int be_suspend(struct pci_dev
*pdev
, pm_message_t state
)
5409 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5411 if (adapter
->wol_en
)
5412 be_setup_wol(adapter
, true);
5414 be_intr_set(adapter
, false);
5415 be_cancel_err_detection(adapter
);
5417 be_cleanup(adapter
);
5419 pci_save_state(pdev
);
5420 pci_disable_device(pdev
);
5421 pci_set_power_state(pdev
, pci_choose_state(pdev
, state
));
5425 static int be_pci_resume(struct pci_dev
*pdev
)
5427 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5430 status
= pci_enable_device(pdev
);
5434 pci_restore_state(pdev
);
5436 status
= be_resume(adapter
);
5440 be_schedule_err_detection(adapter
, ERR_DETECTION_DELAY
);
5442 if (adapter
->wol_en
)
5443 be_setup_wol(adapter
, false);
5449 * An FLR will stop BE from DMAing any data.
5451 static void be_shutdown(struct pci_dev
*pdev
)
5453 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5458 be_roce_dev_shutdown(adapter
);
5459 cancel_delayed_work_sync(&adapter
->work
);
5460 be_cancel_err_detection(adapter
);
5462 netif_device_detach(adapter
->netdev
);
5464 be_cmd_reset_function(adapter
);
5466 pci_disable_device(pdev
);
5469 static pci_ers_result_t
be_eeh_err_detected(struct pci_dev
*pdev
,
5470 pci_channel_state_t state
)
5472 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5474 dev_err(&adapter
->pdev
->dev
, "EEH error detected\n");
5476 be_roce_dev_remove(adapter
);
5478 if (!be_check_error(adapter
, BE_ERROR_EEH
)) {
5479 be_set_error(adapter
, BE_ERROR_EEH
);
5481 be_cancel_err_detection(adapter
);
5483 be_cleanup(adapter
);
5486 if (state
== pci_channel_io_perm_failure
)
5487 return PCI_ERS_RESULT_DISCONNECT
;
5489 pci_disable_device(pdev
);
5491 /* The error could cause the FW to trigger a flash debug dump.
5492 * Resetting the card while flash dump is in progress
5493 * can cause it not to recover; wait for it to finish.
5494 * Wait only for first function as it is needed only once per
5497 if (pdev
->devfn
== 0)
5500 return PCI_ERS_RESULT_NEED_RESET
;
5503 static pci_ers_result_t
be_eeh_reset(struct pci_dev
*pdev
)
5505 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5508 dev_info(&adapter
->pdev
->dev
, "EEH reset\n");
5510 status
= pci_enable_device(pdev
);
5512 return PCI_ERS_RESULT_DISCONNECT
;
5514 pci_set_master(pdev
);
5515 pci_restore_state(pdev
);
5517 /* Check if card is ok and fw is ready */
5518 dev_info(&adapter
->pdev
->dev
,
5519 "Waiting for FW to be ready after EEH reset\n");
5520 status
= be_fw_wait_ready(adapter
);
5522 return PCI_ERS_RESULT_DISCONNECT
;
5524 pci_cleanup_aer_uncorrect_error_status(pdev
);
5525 be_clear_error(adapter
, BE_CLEAR_ALL
);
5526 return PCI_ERS_RESULT_RECOVERED
;
5529 static void be_eeh_resume(struct pci_dev
*pdev
)
5532 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5534 dev_info(&adapter
->pdev
->dev
, "EEH resume\n");
5536 pci_save_state(pdev
);
5538 status
= be_resume(adapter
);
5542 be_roce_dev_add(adapter
);
5544 be_schedule_err_detection(adapter
, ERR_DETECTION_DELAY
);
5547 dev_err(&adapter
->pdev
->dev
, "EEH resume failed\n");
5550 static int be_pci_sriov_configure(struct pci_dev
*pdev
, int num_vfs
)
5552 struct be_adapter
*adapter
= pci_get_drvdata(pdev
);
5557 be_vf_clear(adapter
);
5559 adapter
->num_vfs
= num_vfs
;
5561 if (adapter
->num_vfs
== 0 && pci_vfs_assigned(pdev
)) {
5562 dev_warn(&pdev
->dev
,
5563 "Cannot disable VFs while they are assigned\n");
5567 /* When the HW is in SRIOV capable configuration, the PF-pool resources
5568 * are equally distributed across the max-number of VFs. The user may
5569 * request only a subset of the max-vfs to be enabled.
5570 * Based on num_vfs, redistribute the resources across num_vfs so that
5571 * each VF will have access to more number of resources.
5572 * This facility is not available in BE3 FW.
5573 * Also, this is done by FW in Lancer chip.
5575 if (skyhawk_chip(adapter
) && !pci_num_vf(pdev
)) {
5576 num_vf_qs
= be_calculate_vf_qs(adapter
, adapter
->num_vfs
);
5577 status
= be_cmd_set_sriov_config(adapter
, adapter
->pool_res
,
5578 adapter
->num_vfs
, num_vf_qs
);
5581 "Failed to optimize SR-IOV resources\n");
5584 status
= be_get_resources(adapter
);
5586 return be_cmd_status(status
);
5588 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
5590 status
= be_update_queues(adapter
);
5593 return be_cmd_status(status
);
5595 if (adapter
->num_vfs
)
5596 status
= be_vf_setup(adapter
);
5599 return adapter
->num_vfs
;
5604 static const struct pci_error_handlers be_eeh_handlers
= {
5605 .error_detected
= be_eeh_err_detected
,
5606 .slot_reset
= be_eeh_reset
,
5607 .resume
= be_eeh_resume
,
5610 static struct pci_driver be_driver
= {
5612 .id_table
= be_dev_ids
,
5614 .remove
= be_remove
,
5615 .suspend
= be_suspend
,
5616 .resume
= be_pci_resume
,
5617 .shutdown
= be_shutdown
,
5618 .sriov_configure
= be_pci_sriov_configure
,
5619 .err_handler
= &be_eeh_handlers
5622 static int __init
be_init_module(void)
5624 if (rx_frag_size
!= 8192 && rx_frag_size
!= 4096 &&
5625 rx_frag_size
!= 2048) {
5626 printk(KERN_WARNING DRV_NAME
5627 " : Module param rx_frag_size must be 2048/4096/8192."
5629 rx_frag_size
= 2048;
5633 pr_info(DRV_NAME
" : Module param num_vfs is obsolete.");
5634 pr_info(DRV_NAME
" : Use sysfs method to enable VFs\n");
5637 return pci_register_driver(&be_driver
);
5639 module_init(be_init_module
);
5641 static void __exit
be_exit_module(void)
5643 pci_unregister_driver(&be_driver
);
5645 module_exit(be_exit_module
);